{"module-version": "20241218.01", "data-type": "default whisper output", "whisper-version": "20231106", "source-file": "thunder_day2_1100_-_zacchaeus_scheffer (1080p).mp4", "detected-languages": {"model": {"name": "small", "hash": "sha256:9ecf779972d90ba49c06d968637d720dd632c55bbf19d441fb42bf17a411e794"}, "data": [["en", 0.9894185662269592], ["ja", 0.0035577849484980106], ["la", 0.000973084126599133], ["nn", 0.0008999728015623987], ["cy", 0.0006281708483584225], ["fr", 0.0005589902866631746], ["es", 0.0005280505283735693], ["pt", 0.0004954429459758103], ["ru", 0.0002990350767504424], ["de", 0.00026967070880346], ["ar", 0.0002449334133416414], ["pl", 0.00020021540694870055], ["ko", 0.00019025107030756772], ["zh", 0.00015416672977153212], ["km", 0.00011983592412434518], ["haw", 0.00011406168050598353], ["ms", 0.000107669759017881], ["id", 0.0001030535131576471], ["it", 9.088778460863978e-05], ["jw", 8.292918937513605e-05], ["tl", 6.93641122779809e-05], ["sn", 6.697141361655667e-05], ["tr", 6.239380309125409e-05], ["mi", 5.165100810700096e-05], ["he", 4.831416663364507e-05], ["sv", 4.612896736944094e-05], ["fi", 4.259389970684424e-05], ["hi", 4.245387026458047e-05], ["sw", 4.195862857159227e-05], ["nl", 3.824907980742864e-05], ["hu", 3.306421058368869e-05], ["uk", 3.2774911232991144e-05], ["ml", 3.218396159354597e-05], ["cs", 3.087930963374674e-05], ["ro", 2.9016146072535776e-05], ["el", 2.578183739387896e-05], ["vi", 2.5643903427408077e-05], ["sl", 2.434502857795451e-05], ["fa", 2.2605923732044175e-05], ["ur", 2.184741970268078e-05], ["br", 1.9690327462740242e-05], ["sa", 1.7179549104184844e-05], ["ta", 1.4363636182679329e-05], ["bn", 1.3386530554271303e-05], ["th", 1.3044483239355031e-05], ["no", 1.1908463420695625e-05], ["si", 1.1576472388696857e-05], ["yo", 1.005688409350114e-05], ["bg", 7.332849236263428e-06], ["gl", 5.332904947863426e-06], ["mn", 4.935989181831246e-06], ["da", 4.552548944047885e-06], ["te", 4.517051365837688e-06], ["sk", 4.311204065743368e-06], ["af", 2.8558542908285744e-06], ["ca", 2.7104440505354432e-06], ["hr", 2.693897386052413e-06], ["bo", 2.5304893824795727e-06], ["ht", 2.261044073748053e-06], ["bs", 1.9541521396604367e-06], ["my", 1.3984941915623494e-06], ["lv", 1.3025903626839863e-06], ["eu", 1.2471009540604427e-06], ["yi", 1.2212816500323243e-06], ["hy", 1.2203862524984288e-06], ["fo", 1.0830909786818665e-06], ["is", 1.054511812981218e-06], ["oc", 8.560032256355044e-07], ["sd", 7.364635052908852e-07], ["lt", 6.581557272511418e-07], ["et", 6.40943198959576e-07], ["sr", 6.408301373994618e-07], ["ne", 6.232826308405492e-07], ["ps", 5.510730147761933e-07], ["mr", 5.381426717576687e-07], ["mt", 5.303156740410486e-07], ["be", 5.218186061028973e-07], ["sq", 4.906218578071275e-07], ["az", 4.6011041376914363e-07], ["pa", 3.916184141417034e-07], ["lo", 3.5815398291560996e-07], ["mk", 3.5655631336339866e-07], ["kn", 3.462146196397953e-07], ["as", 2.3208554011944216e-07], ["kk", 1.0269192785017367e-07], ["ka", 9.281549040451864e-08], ["ln", 4.652062912668953e-08], ["am", 4.1354763169465514e-08], ["gu", 2.7484603037919442e-08], ["so", 1.678511374336722e-08], ["tg", 4.7183763562941294e-09], ["lb", 4.134890652096601e-09], ["su", 3.673168436080232e-09], ["tt", 2.114519448426222e-09], ["ha", 1.6151748827297752e-09], ["ba", 5.124374258258513e-10], ["tk", 4.0925191013840845e-10], ["mg", 1.647913333568951e-10], ["uz", 6.129070861948804e-11]]}, "model": {"name": "small.en", "hash": "sha256:f953ad0fd29cacd07d5a9eda5624af0f6bcf2258be67c92b79389873d91e0872"}, "data": {"text": " Thank you for coming to my talk, sharing knowledge with truly open and decentralized AI. My name is Zacchaeus, or just Zacchaeus, and I will be presenting. As an overview of the talk, I will start with an introduction to machine learning, followed by a description of how it has been used. Then I will describe the first attempts at decentralizing machine learning frameworks. Finally, I will point out some of the major design considerations when developing decentralized machine learning frameworks. The last line, avoid open AI, embrace open AI, is the unofficial subtitle of this talk. It is unofficial and varied at the bottom of my overview page because open AI is a major sponsor of this event. By the end of this talk, I want to convince you that we don't need closed source and centralized AI gatekeepers like open AI, and that a decentralized machine learning framework is not only ethically but technologically superior. From a high level, machine learning can be broken down into three components. The first component is a neural network. A neural network is a function f which takes parameters and gives an output. For illustrative purposes, suppose f is a quadratic function and the parameters are the coefficients a, b, and c. The data is a resource which contains knowledge. Often it is censored data from the neural network, but can also be the personal data like your message history. The last component is training, which is an algorithm that takes a neural network's data and extracts knowledge from the data into parameters for the neural network. An art toy example, it finds the a, b, and c which gives you the correct y for any input x. The current state of things. Machine learning requires a neural network, data, and training algorithm. More data means more knowledge and organizations rarely like to share algorithms. Organizations have begun to hoard as much data as possible to collect the most knowledge. Sometimes organizations do share algorithms and sometimes they do share their neural networks, but they will never share their data. That should tell you how valuable your data is. This includes private personal data. There are many motivations to decentralized machine learning. For one, with so many cameras and other sensors in existence, it's not possible to centralize all the data because of data network constraints. More obviously, there are issues with privacy. Maybe you could steal my password or impersonate me and steal my identity. If you share enough data, such things inevitably become possible. Some might say we shouldn't do away with machine learning and keep everything secret, but machine learning has a great potential for good. Suppose I find out that I have been growing cancer for the last three months. Now my data, including the medical diagnosis, contains knowledge that might help others detect cancer before it is too late. Additionally, as I found out at CVPR this year, the carbon footprint of decentralized frameworks can actually be lower than centralized systems because more powerful GPUs are less power efficient. A lot of their power ends up going into active cooling. This is a plot that was provided by Nicholas Lane from the University of Cambridge and Samsung at AI. He presented this at CVPR 2023, less than a week ago, in the Federated Learning Workshop. Can you say the name one more time? The name of the person? Nicholas Lane. So, Federated Learning is a machine learning framework which allows you to train neural networks without centralizing data. It actually achieves the same performance as centralized machine learning methods. It was first designed to contend with laws that prohibit the unauthorized sharing of medical data. Hospitals trained neural networks locally and communicated those parameters to a central server. The central server then combined and redistributed the resulting parameters. This process is repeated until the combined parameters stop improving. Here we have a diagram. Notice that each line here is a hospital and that's the training happening. So you're converting the data into parameters. The data never leaves the hospital and it is eventually aggregated in a central server. The downside here is that it does still require a central server. The basic idea of decentralized federated learning is that instead of exchanging parameters with a central server, you treat each participant as a node in a graph and share with your graph neighbors. There are many ways to do this. Some suggest blockchain. I think blockchain is overkill. Maybe Holochain is the answer. Here I have an example graph. In this case, instead of hospitals, I want you to think of each node as a person. Each node shares knowledge with adjacent nodes and has a certain degree of trust to the knowledge it obtains from its neighbors. I think I train a neural network locally on my compute using my data collected by my sensors. This could be my phone's GPS, my fitness wearable, accelerometer, camera, security system, etc. And then I share that knowledge with others in the network. Unfortunately, the same math that allows you to turn data into parameters can turn parameters into data, your private data. This can be mitigated in several ways, including just sharing parameters less frequently. This is an ongoing area of research, but in general, this problem can be dealt with. It's a losing game. Yes, there's a lot of research that's been done into this. And in general, it is a losing game for the attackers. The ways that you can extract it that have been done, it's very contrived, very... not realistic scenarios. But talk to me after this, if you want to discuss this more, that the next one is a bit more of a problem. The other major consideration is malicious actors. For instance, I could take a bunch of photos of myself, label them as not a burglar, contribute to a federated learning network, and suddenly people's home security system labels me as not a burglar because I lied to the network and told it that things that look like me are not burglars. Basically, we need to develop mechanisms for detecting and ignoring liars. This is also a big area of research. I think it's certainly this is the major hard point of decentralized federated learning, but I do think that it's not an obstacle that cannot be overcome. I think we can overcome this obstacle. So in conclusion, internet networks cannot handle the data throughput to centralize all data. That means more data is available for federated learning frameworks, and so better machine learning models are possible. Federated learning is superior because it uses that data. Organizations want to take your resources, in this case data, and sell you the value, in this case the knowledge stored therein. Free neural networks like LAMA have largely closed the gaps on large language models like chat GPT-3. GPT-4, I know, is a leg up, but I do trust that we are making strides and that we will eventually outperform them simply because there's more data. We don't need centralized AI gatekeepers to control AI and protect us from ourselves. I'll end my talk with an updated subtitle, and here I mean free as in freedom. Avoid open AI, embrace free AI. Thank you. Any questions? We've got one minute for questions. I have a question. So the kind of framing of the threat model here is companies want to take your data so that they can train models off of it and sell it back to you. How does federated learning change that? Now they don't need to take the data, but they still are training a model without me that they can sell back to me. But they're not training it without you. So every node has to have access to the model weights in order for you to converge on one set of model weights. So the community owns the model together. So you're not being sold it. You put in your work. You said, I have this data. I want valuable tools. Maybe it's like I have a voice recognizer. I have a different accent, so I need to add some data for my accent so that it can work. And so I contribute to learn the shared model and I get something out of it and the overall global model benefits as well. So when it's federated, you are always getting a copy of the model locally. Right. Right. At that moment in time. Yes. That's the whole training loop is you get a model parameters. As far as we know, it's the best model weights or parameters, and you tune them a little bit more to improve them just a little bit. And then you send those back out into the world and you're constantly doing that. And if you do that enough, you converge to a better model because there's more data. Question in the back. Is there a something here that we're like essentially averaging the model weights across all of the individuals? And if so, is that appropriate? So the early works on federated learning do literally just average the weights. There are more advanced schemes, especially ones that can help with data privacy. And so sometimes it's a bit more complicated, but at a high level, you should think of it that way is that you're just averaging the parameters. There's some wrinkles in there, but that's how you should think of it. Great. Thank you very much, Zach.", "segments": [{"id": 0, "seek": 0, "start": 1.260000000000005, "end": 6.28, "text": " Thank you for coming to my talk, sharing knowledge with truly open and decentralized AI.", "tokens": [50363, 6952, 345, 329, 2406, 284, 616, 1561, 11, 7373, 3725, 351, 4988, 1280, 290, 26512, 9552, 13, 50720], "temperature": 0.0, "avg_logprob": -0.17923547956678604, "compression_ratio": 1.613733905579399, "no_speech_prob": 0.04298223555088043, "words": [{"word": " Thank", "start": 1.260000000000005, "end": 1.82, "probability": 0.6594257950782776}, {"word": " you", "start": 1.82, "end": 2.0, "probability": 0.9974272847175598}, {"word": " for", "start": 2.0, "end": 2.16, "probability": 0.9871329069137573}, {"word": " coming", "start": 2.16, "end": 2.38, "probability": 0.9930378794670105}, {"word": " to", "start": 2.38, "end": 2.56, "probability": 0.9906055331230164}, {"word": " my", "start": 2.56, "end": 2.7, "probability": 0.995044469833374}, {"word": " talk,", "start": 2.7, "end": 3.04, "probability": 0.9945805668830872}, {"word": " sharing", "start": 3.42, "end": 3.56, "probability": 0.5149276256561279}, {"word": " knowledge", "start": 3.56, "end": 4.0, "probability": 0.9680473208427429}, {"word": " with", "start": 4.0, "end": 4.34, "probability": 0.9943804144859314}, {"word": " truly", "start": 4.34, "end": 4.68, "probability": 0.9853979349136353}, {"word": " open", "start": 4.68, "end": 5.14, "probability": 0.9775741696357727}, {"word": " and", "start": 5.14, "end": 5.36, "probability": 0.9877267479896545}, {"word": " decentralized", "start": 5.36, "end": 5.88, "probability": 0.9862015843391418}, {"word": " AI.", "start": 5.88, "end": 6.28, "probability": 0.9544981718063354}]}, {"id": 1, "seek": 0, "start": 7.12, "end": 10.42, "text": " My name is Zacchaeus, or just Zacchaeus, and I will be presenting.", "tokens": [50720, 2011, 1438, 318, 44922, 354, 46052, 11, 393, 655, 44922, 354, 46052, 11, 290, 314, 481, 307, 17728, 13, 50999], "temperature": 0.0, "avg_logprob": -0.17923547956678604, "compression_ratio": 1.613733905579399, "no_speech_prob": 0.04298223555088043, "words": [{"word": " My", "start": 7.12, "end": 7.24, "probability": 0.9912238121032715}, {"word": " name", "start": 7.24, "end": 7.4, "probability": 0.9992414712905884}, {"word": " is", "start": 7.4, "end": 7.6, "probability": 0.9966670870780945}, {"word": " Zacchaeus,", "start": 7.6, "end": 8.06, "probability": 0.6797821919123331}, {"word": " or", "start": 8.26, "end": 8.38, "probability": 0.9681783318519592}, {"word": " just", "start": 8.38, "end": 8.62, "probability": 0.7069334387779236}, {"word": " Zacchaeus,", "start": 8.62, "end": 9.08, "probability": 0.433398832877477}, {"word": " and", "start": 9.36, "end": 9.6, "probability": 0.9758964776992798}, {"word": " I", "start": 9.6, "end": 9.84, "probability": 0.995120108127594}, {"word": " will", "start": 9.84, "end": 9.94, "probability": 0.9662966728210449}, {"word": " be", "start": 9.94, "end": 10.08, "probability": 0.9974945783615112}, {"word": " presenting.", "start": 10.08, "end": 10.42, "probability": 0.9963357448577881}]}, {"id": 2, "seek": 0, "start": 12.72, "end": 16.62, "text": " As an overview of the talk, I will start with an introduction to machine learning, followed", "tokens": [50999, 1081, 281, 16700, 286, 262, 1561, 11, 314, 481, 923, 351, 281, 9793, 284, 4572, 4673, 11, 3940, 51199], "temperature": 0.0, "avg_logprob": -0.17923547956678604, "compression_ratio": 1.613733905579399, "no_speech_prob": 0.04298223555088043, "words": [{"word": " As", "start": 12.460000000000003, "end": 12.74, "probability": 0.9668812155723572}, {"word": " an", "start": 12.74, "end": 12.86, "probability": 0.997076153755188}, {"word": " overview", "start": 12.86, "end": 13.12, "probability": 0.9988658428192139}, {"word": " of", "start": 13.12, "end": 13.36, "probability": 0.9984416365623474}, {"word": " the", "start": 13.36, "end": 13.48, "probability": 0.980830192565918}, {"word": " talk,", "start": 13.48, "end": 13.76, "probability": 0.9979602098464966}, {"word": " I", "start": 14.06, "end": 14.16, "probability": 0.997374415397644}, {"word": " will", "start": 14.16, "end": 14.26, "probability": 0.9948205947875977}, {"word": " start", "start": 14.26, "end": 14.5, "probability": 0.998666524887085}, {"word": " with", "start": 14.5, "end": 14.7, "probability": 0.9979084730148315}, {"word": " an", "start": 14.7, "end": 14.84, "probability": 0.9957016110420227}, {"word": " introduction", "start": 14.84, "end": 15.18, "probability": 0.9916959404945374}, {"word": " to", "start": 15.18, "end": 15.52, "probability": 0.9975152015686035}, {"word": " machine", "start": 15.52, "end": 15.74, "probability": 0.9564439058303833}, {"word": " learning,", "start": 15.74, "end": 16.06, "probability": 0.9935978651046753}, {"word": " followed", "start": 16.28, "end": 16.62, "probability": 0.9926884770393372}]}, {"id": 3, "seek": 0, "start": 16.62, "end": 18.72, "text": " by a description of how it has been used.", "tokens": [51199, 416, 257, 6764, 286, 703, 340, 468, 587, 973, 13, 51331], "temperature": 0.0, "avg_logprob": -0.17923547956678604, "compression_ratio": 1.613733905579399, "no_speech_prob": 0.04298223555088043, "words": [{"word": " by", "start": 16.62, "end": 16.92, "probability": 0.998891294002533}, {"word": " a", "start": 16.92, "end": 17.04, "probability": 0.9320270419120789}, {"word": " description", "start": 17.04, "end": 17.34, "probability": 0.9990641474723816}, {"word": " of", "start": 17.34, "end": 17.72, "probability": 0.9950540065765381}, {"word": " how", "start": 17.72, "end": 17.94, "probability": 0.999351441860199}, {"word": " it", "start": 17.94, "end": 18.08, "probability": 0.9984908103942871}, {"word": " has", "start": 18.08, "end": 18.2, "probability": 0.9685490131378174}, {"word": " been", "start": 18.2, "end": 18.34, "probability": 0.9992443323135376}, {"word": " used.", "start": 18.34, "end": 18.72, "probability": 0.9986263513565063}]}, {"id": 4, "seek": 0, "start": 19.14, "end": 24.2, "text": " Then I will describe the first attempts at decentralizing machine learning frameworks.", "tokens": [51331, 3244, 314, 481, 6901, 262, 717, 6370, 379, 20649, 2890, 4572, 4673, 29251, 13, 51595], "temperature": 0.0, "avg_logprob": -0.17923547956678604, "compression_ratio": 1.613733905579399, "no_speech_prob": 0.04298223555088043, "words": [{"word": " Then", "start": 19.14, "end": 19.4, "probability": 0.9949722290039062}, {"word": " I", "start": 19.4, "end": 19.7, "probability": 0.7240321636199951}, {"word": " will", "start": 19.7, "end": 19.82, "probability": 0.998146653175354}, {"word": " describe", "start": 19.82, "end": 20.24, "probability": 0.9987320303916931}, {"word": " the", "start": 20.24, "end": 20.54, "probability": 0.9989702701568604}, {"word": " first", "start": 20.54, "end": 20.8, "probability": 0.9979946613311768}, {"word": " attempts", "start": 20.8, "end": 21.42, "probability": 0.9962115287780762}, {"word": " at", "start": 21.42, "end": 22.16, "probability": 0.9976056814193726}, {"word": " decentralizing", "start": 22.16, "end": 23.16, "probability": 0.9578636288642883}, {"word": " machine", "start": 23.16, "end": 23.46, "probability": 0.964570164680481}, {"word": " learning", "start": 23.46, "end": 23.74, "probability": 0.9664663672447205}, {"word": " frameworks.", "start": 23.74, "end": 24.2, "probability": 0.997408926486969}]}, {"id": 5, "seek": 2420, "start": 24.84, "end": 29.28, "text": " Finally, I will point out some of the major design considerations when developing", "tokens": [50395, 9461, 11, 314, 481, 966, 503, 617, 286, 262, 1688, 1486, 18506, 618, 5922, 50643], "temperature": 0.0, "avg_logprob": -0.21652729657231545, "compression_ratio": 1.6704980842911878, "no_speech_prob": 0.22683577239513397, "words": [{"word": " Finally,", "start": 24.779999999999998, "end": 25.08, "probability": 0.35671672224998474}, {"word": " I", "start": 25.5, "end": 25.6, "probability": 0.9756451845169067}, {"word": " will", "start": 25.6, "end": 25.74, "probability": 0.9687245488166809}, {"word": " point", "start": 25.74, "end": 25.92, "probability": 0.9945554733276367}, {"word": " out", "start": 25.92, "end": 26.12, "probability": 0.994706928730011}, {"word": " some", "start": 26.12, "end": 26.26, "probability": 0.982658326625824}, {"word": " of", "start": 26.26, "end": 26.34, "probability": 0.9730101823806763}, {"word": " the", "start": 26.34, "end": 26.46, "probability": 0.9963208436965942}, {"word": " major", "start": 26.46, "end": 26.74, "probability": 0.9914107322692871}, {"word": " design", "start": 26.74, "end": 27.18, "probability": 0.9949357509613037}, {"word": " considerations", "start": 27.18, "end": 28.3, "probability": 0.9892135262489319}, {"word": " when", "start": 28.3, "end": 28.68, "probability": 0.9577197432518005}, {"word": " developing", "start": 28.68, "end": 29.28, "probability": 0.9892072677612305}]}, {"id": 6, "seek": 2420, "start": 29.8, "end": 31.54, "text": " decentralized machine learning frameworks.", "tokens": [50643, 26512, 4572, 4673, 29251, 13, 50779], "temperature": 0.0, "avg_logprob": -0.21652729657231545, "compression_ratio": 1.6704980842911878, "no_speech_prob": 0.22683577239513397, "words": [{"word": " decentralized", "start": 29.8, "end": 30.34, "probability": 0.6716408729553223}, {"word": " machine", "start": 30.34, "end": 30.78, "probability": 0.824600338935852}, {"word": " learning", "start": 30.78, "end": 31.02, "probability": 0.933923065662384}, {"word": " frameworks.", "start": 31.02, "end": 31.54, "probability": 0.9872080683708191}]}, {"id": 7, "seek": 2420, "start": 32.38, "end": 39.5, "text": " The last line, avoid open AI, embrace open AI, is the unofficial subtitle of this talk.", "tokens": [50779, 383, 938, 1627, 11, 3368, 1280, 9552, 11, 12553, 1280, 9552, 11, 318, 262, 27220, 37960, 286, 428, 1561, 13, 51151], "temperature": 0.0, "avg_logprob": -0.21652729657231545, "compression_ratio": 1.6704980842911878, "no_speech_prob": 0.22683577239513397, "words": [{"word": " The", "start": 32.38, "end": 32.68, "probability": 0.10312189161777496}, {"word": " last", "start": 32.68, "end": 32.92, "probability": 0.993923008441925}, {"word": " line,", "start": 32.92, "end": 33.48, "probability": 0.9934051632881165}, {"word": " avoid", "start": 34.38, "end": 34.52, "probability": 0.21342118084430695}, {"word": " open", "start": 34.52, "end": 34.94, "probability": 0.32302382588386536}, {"word": " AI,", "start": 34.94, "end": 35.3, "probability": 0.8992233276367188}, {"word": " embrace", "start": 35.64, "end": 35.88, "probability": 0.9572457671165466}, {"word": " open", "start": 35.88, "end": 36.36, "probability": 0.9754263162612915}, {"word": " AI,", "start": 36.36, "end": 36.9, "probability": 0.9870952367782593}, {"word": " is", "start": 37.32, "end": 37.62, "probability": 0.9827818274497986}, {"word": " the", "start": 37.62, "end": 37.86, "probability": 0.9966877102851868}, {"word": " unofficial", "start": 37.86, "end": 38.32, "probability": 0.9923619031906128}, {"word": " subtitle", "start": 38.32, "end": 38.84, "probability": 0.977222740650177}, {"word": " of", "start": 38.84, "end": 39.08, "probability": 0.9974279999732971}, {"word": " this", "start": 39.08, "end": 39.22, "probability": 0.9946566820144653}, {"word": " talk.", "start": 39.22, "end": 39.5, "probability": 0.9944497346878052}]}, {"id": 8, "seek": 2420, "start": 39.98, "end": 46.66, "text": " It is unofficial and varied at the bottom of my overview page because open AI is a major sponsor of this event.", "tokens": [51151, 632, 318, 27220, 290, 15641, 379, 262, 4220, 286, 616, 16700, 2443, 780, 1280, 9552, 318, 257, 1688, 17487, 286, 428, 1785, 13, 51547], "temperature": 0.0, "avg_logprob": -0.21652729657231545, "compression_ratio": 1.6704980842911878, "no_speech_prob": 0.22683577239513397, "words": [{"word": " It", "start": 39.98, "end": 40.06, "probability": 0.9876477718353271}, {"word": " is", "start": 40.06, "end": 40.16, "probability": 0.9790050983428955}, {"word": " unofficial", "start": 40.16, "end": 40.56, "probability": 0.9951686859130859}, {"word": " and", "start": 40.56, "end": 40.94, "probability": 0.9565895199775696}, {"word": " varied", "start": 40.94, "end": 41.12, "probability": 0.35100868344306946}, {"word": " at", "start": 41.12, "end": 41.28, "probability": 0.9967777132987976}, {"word": " the", "start": 41.28, "end": 41.42, "probability": 0.9994388222694397}, {"word": " bottom", "start": 41.42, "end": 41.66, "probability": 0.9986985921859741}, {"word": " of", "start": 41.66, "end": 42.0, "probability": 0.9981992840766907}, {"word": " my", "start": 42.0, "end": 42.22, "probability": 0.9966339468955994}, {"word": " overview", "start": 42.22, "end": 42.56, "probability": 0.9722723364830017}, {"word": " page", "start": 42.56, "end": 42.98, "probability": 0.9980199337005615}, {"word": " because", "start": 42.98, "end": 43.48, "probability": 0.7215904593467712}, {"word": " open", "start": 43.48, "end": 43.72, "probability": 0.9408932328224182}, {"word": " AI", "start": 43.72, "end": 44.02, "probability": 0.9886975288391113}, {"word": " is", "start": 44.02, "end": 44.42, "probability": 0.9985253214836121}, {"word": " a", "start": 44.42, "end": 44.62, "probability": 0.9967957139015198}, {"word": " major", "start": 44.62, "end": 45.16, "probability": 0.9981243014335632}, {"word": " sponsor", "start": 45.16, "end": 45.82, "probability": 0.998935878276825}, {"word": " of", "start": 45.82, "end": 46.22, "probability": 0.9985178112983704}, {"word": " this", "start": 46.22, "end": 46.34, "probability": 0.998540997505188}, {"word": " event.", "start": 46.34, "end": 46.66, "probability": 0.9928856492042542}]}, {"id": 9, "seek": 2420, "start": 47.94, "end": 54.02, "text": " By the end of this talk, I want to convince you that we don't need closed source and centralized AI gatekeepers", "tokens": [51547, 2750, 262, 886, 286, 428, 1561, 11, 314, 765, 284, 11508, 345, 326, 356, 836, 470, 761, 4838, 2723, 290, 29024, 9552, 8946, 24952, 51855], "temperature": 0.0, "avg_logprob": -0.21652729657231545, "compression_ratio": 1.6704980842911878, "no_speech_prob": 0.22683577239513397, "words": [{"word": " By", "start": 47.94, "end": 48.12, "probability": 0.9308367967605591}, {"word": " the", "start": 48.12, "end": 48.24, "probability": 0.999047577381134}, {"word": " end", "start": 48.24, "end": 48.34, "probability": 0.9995417594909668}, {"word": " of", "start": 48.34, "end": 48.44, "probability": 0.9994716048240662}, {"word": " this", "start": 48.44, "end": 48.58, "probability": 0.9954447746276855}, {"word": " talk,", "start": 48.58, "end": 48.84, "probability": 0.9989462494850159}, {"word": " I", "start": 49.02, "end": 49.18, "probability": 0.9989975094795227}, {"word": " want", "start": 49.18, "end": 49.32, "probability": 0.995624840259552}, {"word": " to", "start": 49.32, "end": 49.48, "probability": 0.9992926120758057}, {"word": " convince", "start": 49.48, "end": 49.76, "probability": 0.9994694590568542}, {"word": " you", "start": 49.76, "end": 50.08, "probability": 0.9985281229019165}, {"word": " that", "start": 50.08, "end": 50.36, "probability": 0.9883450269699097}, {"word": " we", "start": 50.36, "end": 50.5, "probability": 0.9987803101539612}, {"word": " don't", "start": 50.5, "end": 50.66, "probability": 0.9951985776424408}, {"word": " need", "start": 50.66, "end": 51.12, "probability": 0.9992095232009888}, {"word": " closed", "start": 51.12, "end": 51.5, "probability": 0.9599118232727051}, {"word": " source", "start": 51.5, "end": 52.04, "probability": 0.767907440662384}, {"word": " and", "start": 52.04, "end": 52.36, "probability": 0.99151611328125}, {"word": " centralized", "start": 52.36, "end": 52.9, "probability": 0.9939841628074646}, {"word": " AI", "start": 52.9, "end": 53.42, "probability": 0.9786116480827332}, {"word": " gatekeepers", "start": 53.42, "end": 54.02, "probability": 0.9879620373249054}]}, {"id": 10, "seek": 5402, "start": 54.02, "end": 60.12, "text": " like open AI, and that a decentralized machine learning framework is not only ethically but technologically superior.", "tokens": [50363, 588, 1280, 9552, 11, 290, 326, 257, 26512, 4572, 4673, 9355, 318, 407, 691, 4555, 1146, 475, 48299, 9098, 13, 50747], "temperature": 0.0, "avg_logprob": -0.19674270293291876, "compression_ratio": 1.5728643216080402, "no_speech_prob": 0.07129894196987152, "words": [{"word": " like", "start": 54.02, "end": 54.3, "probability": 0.746421217918396}, {"word": " open", "start": 54.3, "end": 54.58, "probability": 0.18978369235992432}, {"word": " AI,", "start": 54.58, "end": 54.86, "probability": 0.5691997408866882}, {"word": " and", "start": 55.24, "end": 55.36, "probability": 0.9799405932426453}, {"word": " that", "start": 55.36, "end": 55.46, "probability": 0.9858453273773193}, {"word": " a", "start": 55.46, "end": 55.58, "probability": 0.9795536994934082}, {"word": " decentralized", "start": 55.58, "end": 56.14, "probability": 0.9848095774650574}, {"word": " machine", "start": 56.14, "end": 56.54, "probability": 0.8521414399147034}, {"word": " learning", "start": 56.54, "end": 56.8, "probability": 0.9422212243080139}, {"word": " framework", "start": 56.8, "end": 57.24, "probability": 0.9979169964790344}, {"word": " is", "start": 57.24, "end": 57.52, "probability": 0.9884051084518433}, {"word": " not", "start": 57.52, "end": 57.68, "probability": 0.9929226636886597}, {"word": " only", "start": 57.68, "end": 57.96, "probability": 0.9943636655807495}, {"word": " ethically", "start": 57.96, "end": 58.54, "probability": 0.9952123761177063}, {"word": " but", "start": 58.54, "end": 58.96, "probability": 0.6004267930984497}, {"word": " technologically", "start": 58.96, "end": 59.52, "probability": 0.9933890104293823}, {"word": " superior.", "start": 59.52, "end": 60.12, "probability": 0.9973153471946716}]}, {"id": 11, "seek": 5402, "start": 61.88, "end": 66.4, "text": " From a high level, machine learning can be broken down into three components.", "tokens": [50747, 3574, 257, 1029, 1241, 11, 4572, 4673, 460, 307, 5445, 866, 656, 1115, 6805, 13, 50995], "temperature": 0.0, "avg_logprob": -0.19674270293291876, "compression_ratio": 1.5728643216080402, "no_speech_prob": 0.07129894196987152, "words": [{"word": " From", "start": 61.88, "end": 61.98, "probability": 0.9932180047035217}, {"word": " a", "start": 61.98, "end": 62.08, "probability": 0.9896426796913147}, {"word": " high", "start": 62.08, "end": 62.22, "probability": 0.9969283938407898}, {"word": " level,", "start": 62.22, "end": 62.52, "probability": 0.966308057308197}, {"word": " machine", "start": 63.32, "end": 63.42, "probability": 0.9520758390426636}, {"word": " learning", "start": 63.42, "end": 63.72, "probability": 0.9970332384109497}, {"word": " can", "start": 63.72, "end": 64.1, "probability": 0.993645429611206}, {"word": " be", "start": 64.1, "end": 64.3, "probability": 0.9948478937149048}, {"word": " broken", "start": 64.3, "end": 64.7, "probability": 0.9974925518035889}, {"word": " down", "start": 64.7, "end": 65.1, "probability": 0.9939855933189392}, {"word": " into", "start": 65.1, "end": 65.64, "probability": 0.9906288981437683}, {"word": " three", "start": 65.64, "end": 65.92, "probability": 0.9314072728157043}, {"word": " components.", "start": 65.92, "end": 66.4, "probability": 0.9954256415367126}]}, {"id": 12, "seek": 5402, "start": 66.8, "end": 68.68, "text": " The first component is a neural network.", "tokens": [50995, 383, 717, 7515, 318, 257, 17019, 3127, 13, 51107], "temperature": 0.0, "avg_logprob": -0.19674270293291876, "compression_ratio": 1.5728643216080402, "no_speech_prob": 0.07129894196987152, "words": [{"word": " The", "start": 66.8, "end": 66.96, "probability": 0.9904854893684387}, {"word": " first", "start": 66.96, "end": 67.18, "probability": 0.9985373020172119}, {"word": " component", "start": 67.18, "end": 67.62, "probability": 0.9980175495147705}, {"word": " is", "start": 67.62, "end": 67.98, "probability": 0.9943015575408936}, {"word": " a", "start": 67.98, "end": 68.14, "probability": 0.9880245327949524}, {"word": " neural", "start": 68.14, "end": 68.26, "probability": 0.9927639365196228}, {"word": " network.", "start": 68.26, "end": 68.68, "probability": 0.9987928867340088}]}, {"id": 13, "seek": 5402, "start": 69.22, "end": 79.34, "text": " A neural network is a function f which takes parameters and gives an output.", "tokens": [51107, 317, 17019, 3127, 318, 257, 2163, 277, 543, 2753, 10007, 290, 3607, 281, 5072, 13, 51831], "temperature": 0.0, "avg_logprob": -0.19674270293291876, "compression_ratio": 1.5728643216080402, "no_speech_prob": 0.07129894196987152, "words": [{"word": " A", "start": 69.22, "end": 69.6, "probability": 0.9577328562736511}, {"word": " neural", "start": 69.6, "end": 69.86, "probability": 0.9947878122329712}, {"word": " network", "start": 69.86, "end": 70.28, "probability": 0.9991379976272583}, {"word": " is", "start": 70.28, "end": 71.02, "probability": 0.9907907247543335}, {"word": " a", "start": 71.02, "end": 71.28, "probability": 0.9946320056915283}, {"word": " function", "start": 71.28, "end": 71.6, "probability": 0.9500584006309509}, {"word": " f", "start": 71.6, "end": 72.22, "probability": 0.49905866384506226}, {"word": " which", "start": 72.22, "end": 73.18, "probability": 0.5801145434379578}, {"word": " takes", "start": 73.18, "end": 73.58, "probability": 0.9947722554206848}, {"word": " parameters", "start": 73.58, "end": 74.28, "probability": 0.992068886756897}, {"word": " and", "start": 74.28, "end": 78.08, "probability": 0.374125599861145}, {"word": " gives", "start": 78.08, "end": 78.86, "probability": 0.7179327011108398}, {"word": " an", "start": 78.86, "end": 79.02, "probability": 0.9873588681221008}, {"word": " output.", "start": 79.02, "end": 79.34, "probability": 0.9862318634986877}]}, {"id": 14, "seek": 7934, "start": 83.46000000000001, "end": 90.82, "text": " For illustrative purposes, suppose f is a quadratic function and the parameters are the coefficients a, b, and c.", "tokens": [50411, 1114, 6406, 13260, 4959, 11, 11691, 277, 318, 257, 15094, 81, 1512, 2163, 290, 262, 10007, 389, 262, 44036, 257, 11, 275, 11, 290, 269, 13, 50959], "temperature": 0.0, "avg_logprob": -0.1720037573859805, "compression_ratio": 1.6133333333333333, "no_speech_prob": 0.005361212883144617, "words": [{"word": " For", "start": 83.46000000000001, "end": 84.06, "probability": 0.23210638761520386}, {"word": " illustrative", "start": 84.06, "end": 84.66, "probability": 0.944635272026062}, {"word": " purposes,", "start": 84.66, "end": 85.5, "probability": 0.9885148406028748}, {"word": " suppose", "start": 85.98, "end": 86.02, "probability": 0.9539576172828674}, {"word": " f", "start": 86.02, "end": 86.34, "probability": 0.36868414282798767}, {"word": " is", "start": 86.34, "end": 86.5, "probability": 0.9820441007614136}, {"word": " a", "start": 86.5, "end": 86.68, "probability": 0.9723529815673828}, {"word": " quadratic", "start": 86.68, "end": 87.08, "probability": 0.9947671492894491}, {"word": " function", "start": 87.08, "end": 87.46, "probability": 0.9967969059944153}, {"word": " and", "start": 87.46, "end": 87.78, "probability": 0.6541604399681091}, {"word": " the", "start": 87.78, "end": 87.9, "probability": 0.9641326665878296}, {"word": " parameters", "start": 87.9, "end": 88.26, "probability": 0.950711727142334}, {"word": " are", "start": 88.26, "end": 88.74, "probability": 0.9873212575912476}, {"word": " the", "start": 88.74, "end": 88.96, "probability": 0.9090576767921448}, {"word": " coefficients", "start": 88.96, "end": 89.34, "probability": 0.9692145586013794}, {"word": " a,", "start": 89.34, "end": 89.84, "probability": 0.7397229075431824}, {"word": " b,", "start": 90.08, "end": 90.3, "probability": 0.9528592824935913}, {"word": " and", "start": 90.44, "end": 90.6, "probability": 0.9847480654716492}, {"word": " c.", "start": 90.6, "end": 90.82, "probability": 0.9945698380470276}]}, {"id": 15, "seek": 7934, "start": 91.36, "end": 94.16, "text": " The data is a resource which contains knowledge.", "tokens": [50959, 383, 1366, 318, 257, 8271, 543, 4909, 3725, 13, 51119], "temperature": 0.0, "avg_logprob": -0.1720037573859805, "compression_ratio": 1.6133333333333333, "no_speech_prob": 0.005361212883144617, "words": [{"word": " The", "start": 91.36, "end": 91.58, "probability": 0.989014208316803}, {"word": " data", "start": 91.58, "end": 91.78, "probability": 0.9977527260780334}, {"word": " is", "start": 91.78, "end": 92.1, "probability": 0.997510552406311}, {"word": " a", "start": 92.1, "end": 92.32, "probability": 0.9808875322341919}, {"word": " resource", "start": 92.32, "end": 92.78, "probability": 0.9903916120529175}, {"word": " which", "start": 92.78, "end": 93.18, "probability": 0.9452793002128601}, {"word": " contains", "start": 93.18, "end": 93.68, "probability": 0.9946540594100952}, {"word": " knowledge.", "start": 93.68, "end": 94.16, "probability": 0.99741530418396}]}, {"id": 16, "seek": 7934, "start": 94.64, "end": 101.56, "text": " Often it is censored data from the neural network, but can also be the personal data like your message history.", "tokens": [51119, 18023, 340, 318, 43847, 1366, 422, 262, 17019, 3127, 11, 475, 460, 635, 307, 262, 2614, 1366, 588, 534, 3275, 2106, 13, 51503], "temperature": 0.0, "avg_logprob": -0.1720037573859805, "compression_ratio": 1.6133333333333333, "no_speech_prob": 0.005361212883144617, "words": [{"word": " Often", "start": 94.64, "end": 94.88, "probability": 0.9887906312942505}, {"word": " it", "start": 94.88, "end": 95.12, "probability": 0.6784749031066895}, {"word": " is", "start": 95.12, "end": 95.3, "probability": 0.9918416738510132}, {"word": " censored", "start": 95.3, "end": 95.54, "probability": 0.4914512634277344}, {"word": " data", "start": 95.54, "end": 95.92, "probability": 0.9915087223052979}, {"word": " from", "start": 95.92, "end": 96.52, "probability": 0.9903481006622314}, {"word": " the", "start": 96.52, "end": 96.94, "probability": 0.979331374168396}, {"word": " neural", "start": 96.94, "end": 97.16, "probability": 0.9915087223052979}, {"word": " network,", "start": 97.16, "end": 97.5, "probability": 0.9882184863090515}, {"word": " but", "start": 97.6, "end": 97.8, "probability": 0.9960154891014099}, {"word": " can", "start": 97.8, "end": 97.96, "probability": 0.7878997325897217}, {"word": " also", "start": 97.96, "end": 98.32, "probability": 0.9950655102729797}, {"word": " be", "start": 98.32, "end": 98.68, "probability": 0.9940989017486572}, {"word": " the", "start": 98.68, "end": 99.42, "probability": 0.9661175012588501}, {"word": " personal", "start": 99.42, "end": 99.94, "probability": 0.9915645122528076}, {"word": " data", "start": 99.94, "end": 100.24, "probability": 0.9974748492240906}, {"word": " like", "start": 100.24, "end": 100.56, "probability": 0.7035964727401733}, {"word": " your", "start": 100.56, "end": 100.74, "probability": 0.9848769903182983}, {"word": " message", "start": 100.74, "end": 101.1, "probability": 0.9952154159545898}, {"word": " history.", "start": 101.1, "end": 101.56, "probability": 0.9980765581130981}]}, {"id": 17, "seek": 7934, "start": 102.22, "end": 106.66, "text": " The last component is training, which is an algorithm that takes a neural network's data", "tokens": [51503, 383, 938, 7515, 318, 3047, 11, 543, 318, 281, 11862, 326, 2753, 257, 17019, 3127, 338, 1366, 51743], "temperature": 0.0, "avg_logprob": -0.1720037573859805, "compression_ratio": 1.6133333333333333, "no_speech_prob": 0.005361212883144617, "words": [{"word": " The", "start": 102.22, "end": 102.44, "probability": 0.9945664405822754}, {"word": " last", "start": 102.44, "end": 102.66, "probability": 0.9993777275085449}, {"word": " component", "start": 102.66, "end": 102.96, "probability": 0.9980375170707703}, {"word": " is", "start": 102.96, "end": 103.26, "probability": 0.9983720183372498}, {"word": " training,", "start": 103.26, "end": 103.56, "probability": 0.9765982627868652}, {"word": " which", "start": 103.84, "end": 103.98, "probability": 0.9944063425064087}, {"word": " is", "start": 103.98, "end": 104.14, "probability": 0.9965057373046875}, {"word": " an", "start": 104.14, "end": 104.3, "probability": 0.9957455992698669}, {"word": " algorithm", "start": 104.3, "end": 104.72, "probability": 0.9978779554367065}, {"word": " that", "start": 104.72, "end": 105.18, "probability": 0.9911330342292786}, {"word": " takes", "start": 105.18, "end": 105.48, "probability": 0.9971780776977539}, {"word": " a", "start": 105.48, "end": 105.64, "probability": 0.824770450592041}, {"word": " neural", "start": 105.64, "end": 105.84, "probability": 0.9970996379852295}, {"word": " network's", "start": 105.84, "end": 106.44, "probability": 0.9301692545413971}, {"word": " data", "start": 106.44, "end": 106.66, "probability": 0.9983684420585632}]}, {"id": 18, "seek": 10666, "start": 106.66, "end": 111.44, "text": " and extracts knowledge from the data into parameters for the neural network.", "tokens": [50379, 290, 32139, 3725, 422, 262, 1366, 656, 10007, 329, 262, 17019, 3127, 13, 50599], "temperature": 0.0, "avg_logprob": -0.15261751267968154, "compression_ratio": 1.6192660550458715, "no_speech_prob": 0.006900294218212366, "words": [{"word": " and", "start": 106.66, "end": 107.16, "probability": 0.198451429605484}, {"word": " extracts", "start": 107.16, "end": 107.8, "probability": 0.9524737596511841}, {"word": " knowledge", "start": 107.8, "end": 108.34, "probability": 0.9784278869628906}, {"word": " from", "start": 108.34, "end": 108.8, "probability": 0.9851363897323608}, {"word": " the", "start": 108.8, "end": 108.96, "probability": 0.9691159129142761}, {"word": " data", "start": 108.96, "end": 109.18, "probability": 0.9973949193954468}, {"word": " into", "start": 109.18, "end": 109.54, "probability": 0.9752389788627625}, {"word": " parameters", "start": 109.54, "end": 110.02, "probability": 0.9827828407287598}, {"word": " for", "start": 110.02, "end": 110.64, "probability": 0.9779982566833496}, {"word": " the", "start": 110.64, "end": 110.86, "probability": 0.9859539866447449}, {"word": " neural", "start": 110.86, "end": 111.04, "probability": 0.9702061414718628}, {"word": " network.", "start": 111.04, "end": 111.44, "probability": 0.9904787540435791}]}, {"id": 19, "seek": 10666, "start": 112.04, "end": 121.92, "text": " An art toy example, it finds the a, b, and c which gives you the correct y for any input x.", "tokens": [50647, 1052, 1242, 13373, 1672, 11, 340, 7228, 262, 257, 11, 275, 11, 290, 269, 543, 3607, 345, 262, 3376, 331, 329, 597, 5128, 2124, 13, 51119], "temperature": 0.0, "avg_logprob": -0.15261751267968154, "compression_ratio": 1.6192660550458715, "no_speech_prob": 0.006900294218212366, "words": [{"word": " An", "start": 112.04, "end": 112.62, "probability": 0.3406006693840027}, {"word": " art", "start": 112.62, "end": 112.82, "probability": 0.2751447260379791}, {"word": " toy", "start": 112.82, "end": 113.3, "probability": 0.4667356014251709}, {"word": " example,", "start": 113.3, "end": 115.82, "probability": 0.6727318167686462}, {"word": " it", "start": 116.32, "end": 116.52, "probability": 0.94782555103302}, {"word": " finds", "start": 116.52, "end": 117.02, "probability": 0.986943244934082}, {"word": " the", "start": 117.02, "end": 117.44, "probability": 0.93696129322052}, {"word": " a,", "start": 117.44, "end": 117.68, "probability": 0.15750752389431}, {"word": " b,", "start": 117.74, "end": 117.92, "probability": 0.9728916883468628}, {"word": " and", "start": 117.96, "end": 118.08, "probability": 0.9807776808738708}, {"word": " c", "start": 118.08, "end": 118.34, "probability": 0.9942638278007507}, {"word": " which", "start": 118.34, "end": 118.64, "probability": 0.4729287624359131}, {"word": " gives", "start": 118.64, "end": 118.9, "probability": 0.9721454381942749}, {"word": " you", "start": 118.9, "end": 119.4, "probability": 0.8229430913925171}, {"word": " the", "start": 119.4, "end": 119.74, "probability": 0.9871892333030701}, {"word": " correct", "start": 119.74, "end": 120.1, "probability": 0.9953402280807495}, {"word": " y", "start": 120.1, "end": 120.58, "probability": 0.7238597273826599}, {"word": " for", "start": 120.58, "end": 120.88, "probability": 0.9836527109146118}, {"word": " any", "start": 120.88, "end": 121.1, "probability": 0.9875572919845581}, {"word": " input", "start": 121.1, "end": 121.46, "probability": 0.9991262555122375}, {"word": " x.", "start": 121.46, "end": 121.92, "probability": 0.8107460737228394}]}, {"id": 20, "seek": 10666, "start": 123.66, "end": 124.76, "text": " The current state of things.", "tokens": [51203, 383, 1459, 1181, 286, 1243, 13, 51287], "temperature": 0.0, "avg_logprob": -0.15261751267968154, "compression_ratio": 1.6192660550458715, "no_speech_prob": 0.006900294218212366, "words": [{"word": " The", "start": 123.66, "end": 123.74, "probability": 0.9210836291313171}, {"word": " current", "start": 123.74, "end": 123.94, "probability": 0.9291293621063232}, {"word": " state", "start": 123.94, "end": 124.2, "probability": 0.9900229573249817}, {"word": " of", "start": 124.2, "end": 124.34, "probability": 0.9972190856933594}, {"word": " things.", "start": 124.34, "end": 124.76, "probability": 0.9948069453239441}]}, {"id": 21, "seek": 10666, "start": 125.58, "end": 129.66, "text": " Machine learning requires a neural network, data, and training algorithm.", "tokens": [51287, 10850, 4673, 4433, 257, 17019, 3127, 11, 1366, 11, 290, 3047, 11862, 13, 51539], "temperature": 0.0, "avg_logprob": -0.15261751267968154, "compression_ratio": 1.6192660550458715, "no_speech_prob": 0.006900294218212366, "words": [{"word": " Machine", "start": 125.58, "end": 125.58, "probability": 0.9725114107131958}, {"word": " learning", "start": 125.58, "end": 126.08, "probability": 0.932268500328064}, {"word": " requires", "start": 126.08, "end": 126.8, "probability": 0.9956703186035156}, {"word": " a", "start": 126.8, "end": 127.08, "probability": 0.9832573533058167}, {"word": " neural", "start": 127.08, "end": 127.3, "probability": 0.9954429864883423}, {"word": " network,", "start": 127.3, "end": 127.68, "probability": 0.9981642365455627}, {"word": " data,", "start": 128.12, "end": 128.4, "probability": 0.9963017702102661}, {"word": " and", "start": 128.66, "end": 128.9, "probability": 0.9967290163040161}, {"word": " training", "start": 128.9, "end": 129.26, "probability": 0.9591853022575378}, {"word": " algorithm.", "start": 129.26, "end": 129.66, "probability": 0.9141072034835815}]}, {"id": 22, "seek": 10666, "start": 130.3, "end": 134.86, "text": " More data means more knowledge and organizations rarely like to share algorithms.", "tokens": [51539, 3125, 1366, 1724, 517, 3725, 290, 5745, 8365, 588, 284, 2648, 16113, 13, 51791], "temperature": 0.0, "avg_logprob": -0.15261751267968154, "compression_ratio": 1.6192660550458715, "no_speech_prob": 0.006900294218212366, "words": [{"word": " More", "start": 130.3, "end": 130.52, "probability": 0.9971461892127991}, {"word": " data", "start": 130.52, "end": 130.84, "probability": 0.9992144107818604}, {"word": " means", "start": 130.84, "end": 131.2, "probability": 0.9870282411575317}, {"word": " more", "start": 131.2, "end": 131.48, "probability": 0.9987452030181885}, {"word": " knowledge", "start": 131.48, "end": 131.96, "probability": 0.9988434314727783}, {"word": " and", "start": 131.96, "end": 132.54, "probability": 0.6098713874816895}, {"word": " organizations", "start": 132.54, "end": 133.06, "probability": 0.9761196374893188}, {"word": " rarely", "start": 133.06, "end": 133.52, "probability": 0.9663333892822266}, {"word": " like", "start": 133.52, "end": 133.86, "probability": 0.991628885269165}, {"word": " to", "start": 133.86, "end": 134.06, "probability": 0.9982826709747314}, {"word": " share", "start": 134.06, "end": 134.28, "probability": 0.9987779259681702}, {"word": " algorithms.", "start": 134.28, "end": 134.86, "probability": 0.9953742623329163}]}, {"id": 23, "seek": 13486, "start": 135.5, "end": 141.34, "text": " Organizations have begun to hoard as much data as possible to collect the most knowledge.", "tokens": [50395, 41846, 423, 9258, 284, 46960, 355, 881, 1366, 355, 1744, 284, 2824, 262, 749, 3725, 13, 50707], "temperature": 0.0, "avg_logprob": -0.10364921371658127, "compression_ratio": 1.654708520179372, "no_speech_prob": 0.010265474207699299, "words": [{"word": " Organizations", "start": 135.5, "end": 135.9, "probability": 0.006052364595234394}, {"word": " have", "start": 135.9, "end": 136.68, "probability": 0.8886415362358093}, {"word": " begun", "start": 136.68, "end": 137.04, "probability": 0.9109653234481812}, {"word": " to", "start": 137.04, "end": 137.34, "probability": 0.9942104816436768}, {"word": " hoard", "start": 137.34, "end": 137.82, "probability": 0.9808933734893799}, {"word": " as", "start": 137.82, "end": 138.34, "probability": 0.973530113697052}, {"word": " much", "start": 138.34, "end": 138.64, "probability": 0.9943020343780518}, {"word": " data", "start": 138.64, "end": 138.9, "probability": 0.9911229610443115}, {"word": " as", "start": 138.9, "end": 139.12, "probability": 0.9786604046821594}, {"word": " possible", "start": 139.12, "end": 139.8, "probability": 0.9893348813056946}, {"word": " to", "start": 139.8, "end": 140.18, "probability": 0.9480484127998352}, {"word": " collect", "start": 140.18, "end": 140.38, "probability": 0.9959730505943298}, {"word": " the", "start": 140.38, "end": 140.56, "probability": 0.9890183210372925}, {"word": " most", "start": 140.56, "end": 140.82, "probability": 0.9971290230751038}, {"word": " knowledge.", "start": 140.82, "end": 141.34, "probability": 0.9962608218193054}]}, {"id": 24, "seek": 13486, "start": 141.92, "end": 147.32, "text": " Sometimes organizations do share algorithms and sometimes they do share their neural networks,", "tokens": [50707, 8975, 5745, 466, 2648, 16113, 290, 3360, 484, 466, 2648, 511, 17019, 7686, 11, 50983], "temperature": 0.0, "avg_logprob": -0.10364921371658127, "compression_ratio": 1.654708520179372, "no_speech_prob": 0.010265474207699299, "words": [{"word": " Sometimes", "start": 141.92, "end": 142.22, "probability": 0.9473695755004883}, {"word": " organizations", "start": 142.22, "end": 142.76, "probability": 0.8547379970550537}, {"word": " do", "start": 142.76, "end": 143.36, "probability": 0.9157736301422119}, {"word": " share", "start": 143.36, "end": 144.06, "probability": 0.9972056746482849}, {"word": " algorithms", "start": 144.06, "end": 144.64, "probability": 0.9903338551521301}, {"word": " and", "start": 144.64, "end": 145.24, "probability": 0.5611764788627625}, {"word": " sometimes", "start": 145.24, "end": 145.58, "probability": 0.9702220559120178}, {"word": " they", "start": 145.58, "end": 145.8, "probability": 0.9489271640777588}, {"word": " do", "start": 145.8, "end": 146.08, "probability": 0.9233842492103577}, {"word": " share", "start": 146.08, "end": 146.48, "probability": 0.9964095950126648}, {"word": " their", "start": 146.48, "end": 146.72, "probability": 0.9838671684265137}, {"word": " neural", "start": 146.72, "end": 147.0, "probability": 0.9871925115585327}, {"word": " networks,", "start": 147.0, "end": 147.32, "probability": 0.9961147308349609}]}, {"id": 25, "seek": 13486, "start": 147.46, "end": 149.18, "text": " but they will never share their data.", "tokens": [50983, 475, 484, 481, 1239, 2648, 511, 1366, 13, 51079], "temperature": 0.0, "avg_logprob": -0.10364921371658127, "compression_ratio": 1.654708520179372, "no_speech_prob": 0.010265474207699299, "words": [{"word": " but", "start": 147.46, "end": 147.54, "probability": 0.9966381788253784}, {"word": " they", "start": 147.54, "end": 147.66, "probability": 0.9946145415306091}, {"word": " will", "start": 147.66, "end": 147.78, "probability": 0.994181215763092}, {"word": " never", "start": 147.78, "end": 148.1, "probability": 0.9945106506347656}, {"word": " share", "start": 148.1, "end": 148.68, "probability": 0.9963209629058838}, {"word": " their", "start": 148.68, "end": 148.88, "probability": 0.9951570630073547}, {"word": " data.", "start": 148.88, "end": 149.18, "probability": 0.9984990358352661}]}, {"id": 26, "seek": 13486, "start": 150.08, "end": 152.54000000000002, "text": " That should tell you how valuable your data is.", "tokens": [51123, 1320, 815, 1560, 345, 703, 8119, 534, 1366, 318, 13, 51247], "temperature": 0.0, "avg_logprob": -0.10364921371658127, "compression_ratio": 1.654708520179372, "no_speech_prob": 0.010265474207699299, "words": [{"word": " That", "start": 150.08, "end": 150.34, "probability": 0.9749748110771179}, {"word": " should", "start": 150.34, "end": 150.5, "probability": 0.9951509237289429}, {"word": " tell", "start": 150.5, "end": 150.72, "probability": 0.9980290532112122}, {"word": " you", "start": 150.72, "end": 150.98, "probability": 0.9983440637588501}, {"word": " how", "start": 150.98, "end": 151.26, "probability": 0.9949374198913574}, {"word": " valuable", "start": 151.26, "end": 151.7, "probability": 0.9995443224906921}, {"word": " your", "start": 151.7, "end": 151.96, "probability": 0.9961657524108887}, {"word": " data", "start": 151.96, "end": 152.18, "probability": 0.9991973042488098}, {"word": " is.", "start": 152.18, "end": 152.54000000000002, "probability": 0.9931291341781616}]}, {"id": 27, "seek": 13486, "start": 155.1, "end": 158.24, "text": " This includes private personal data.", "tokens": [51395, 770, 3407, 2839, 2614, 1366, 13, 51535], "temperature": 0.0, "avg_logprob": -0.10364921371658127, "compression_ratio": 1.654708520179372, "no_speech_prob": 0.010265474207699299, "words": [{"word": " This", "start": 155.1, "end": 155.78, "probability": 0.9487731456756592}, {"word": " includes", "start": 155.78, "end": 156.18, "probability": 0.99429851770401}, {"word": " private", "start": 156.18, "end": 156.6, "probability": 0.9831538796424866}, {"word": " personal", "start": 156.6, "end": 157.86, "probability": 0.8209955096244812}, {"word": " data.", "start": 157.86, "end": 158.24, "probability": 0.9991348385810852}]}, {"id": 28, "seek": 13486, "start": 159.34, "end": 162.3, "text": " There are many motivations to decentralized machine learning.", "tokens": [51579, 1318, 389, 867, 28140, 284, 26512, 4572, 4673, 13, 51743], "temperature": 0.0, "avg_logprob": -0.10364921371658127, "compression_ratio": 1.654708520179372, "no_speech_prob": 0.010265474207699299, "words": [{"word": " There", "start": 159.34, "end": 159.5, "probability": 0.9914907813072205}, {"word": " are", "start": 159.5, "end": 159.58, "probability": 0.9875120520591736}, {"word": " many", "start": 159.58, "end": 159.7, "probability": 0.9981436729431152}, {"word": " motivations", "start": 159.7, "end": 160.24, "probability": 0.9763408303260803}, {"word": " to", "start": 160.24, "end": 160.9, "probability": 0.9910420179367065}, {"word": " decentralized", "start": 160.9, "end": 161.56, "probability": 0.4800555109977722}, {"word": " machine", "start": 161.56, "end": 162.02, "probability": 0.9853604435920715}, {"word": " learning.", "start": 162.02, "end": 162.3, "probability": 0.9858623147010803}]}, {"id": 29, "seek": 16230, "start": 162.3, "end": 168.18, "text": " For one, with so many cameras and other sensors in existence, it's not possible to centralize all the data", "tokens": [50371, 1114, 530, 11, 351, 523, 867, 9073, 290, 584, 15736, 287, 6224, 11, 340, 338, 407, 1744, 284, 4318, 1096, 477, 262, 1366, 50667], "temperature": 0.0, "avg_logprob": -0.12729211287064987, "compression_ratio": 1.5324074074074074, "no_speech_prob": 0.0075220344588160515, "words": [{"word": " For", "start": 162.3, "end": 162.72, "probability": 0.41141194105148315}, {"word": " one,", "start": 162.72, "end": 163.02, "probability": 0.9649704694747925}, {"word": " with", "start": 163.22, "end": 163.38, "probability": 0.9900044798851013}, {"word": " so", "start": 163.38, "end": 163.54, "probability": 0.9910213351249695}, {"word": " many", "start": 163.54, "end": 163.72, "probability": 0.9985137581825256}, {"word": " cameras", "start": 163.72, "end": 164.08, "probability": 0.9932432770729065}, {"word": " and", "start": 164.08, "end": 164.3, "probability": 0.9781211018562317}, {"word": " other", "start": 164.3, "end": 164.54, "probability": 0.9893052577972412}, {"word": " sensors", "start": 164.54, "end": 164.88, "probability": 0.9931095838546753}, {"word": " in", "start": 164.88, "end": 165.12, "probability": 0.9901124835014343}, {"word": " existence,", "start": 165.12, "end": 165.5, "probability": 0.9957669973373413}, {"word": " it's", "start": 165.98, "end": 166.08, "probability": 0.9701650142669678}, {"word": " not", "start": 166.08, "end": 166.26, "probability": 0.9985295534133911}, {"word": " possible", "start": 166.26, "end": 166.92, "probability": 0.9987519979476929}, {"word": " to", "start": 166.92, "end": 167.22, "probability": 0.9966480135917664}, {"word": " centralize", "start": 167.22, "end": 167.64, "probability": 0.9892492890357971}, {"word": " all", "start": 167.64, "end": 167.8, "probability": 0.9900450110435486}, {"word": " the", "start": 167.8, "end": 167.92, "probability": 0.961584210395813}, {"word": " data", "start": 167.92, "end": 168.18, "probability": 0.9994204044342041}]}, {"id": 30, "seek": 16230, "start": 168.18, "end": 171.34, "text": " because of data network constraints.", "tokens": [50667, 780, 286, 1366, 3127, 17778, 13, 50839], "temperature": 0.0, "avg_logprob": -0.12729211287064987, "compression_ratio": 1.5324074074074074, "no_speech_prob": 0.0075220344588160515, "words": [{"word": " because", "start": 168.18, "end": 168.84, "probability": 0.9077646732330322}, {"word": " of", "start": 168.84, "end": 169.42, "probability": 0.9956199526786804}, {"word": " data", "start": 169.42, "end": 170.1, "probability": 0.9814997911453247}, {"word": " network", "start": 170.1, "end": 170.56, "probability": 0.9474162459373474}, {"word": " constraints.", "start": 170.56, "end": 171.34, "probability": 0.9781061410903931}]}, {"id": 31, "seek": 16230, "start": 172.06, "end": 176.04, "text": " More obviously, there are issues with privacy.", "tokens": [50839, 3125, 6189, 11, 612, 389, 2428, 351, 6782, 13, 51059], "temperature": 0.0, "avg_logprob": -0.12729211287064987, "compression_ratio": 1.5324074074074074, "no_speech_prob": 0.0075220344588160515, "words": [{"word": " More", "start": 172.06, "end": 172.14, "probability": 0.8984034657478333}, {"word": " obviously,", "start": 172.14, "end": 172.54, "probability": 0.9961283206939697}, {"word": " there", "start": 173.06, "end": 173.14, "probability": 0.9302360415458679}, {"word": " are", "start": 173.14, "end": 173.22, "probability": 0.9730963110923767}, {"word": " issues", "start": 173.22, "end": 173.52, "probability": 0.9353805780410767}, {"word": " with", "start": 173.52, "end": 174.58, "probability": 0.9787299036979675}, {"word": " privacy.", "start": 174.58, "end": 176.04, "probability": 0.5524416565895081}]}, {"id": 32, "seek": 16230, "start": 177.52, "end": 183.12, "text": " Maybe you could steal my password or impersonate me and steal my identity.", "tokens": [51123, 6674, 345, 714, 8711, 616, 9206, 393, 28671, 378, 502, 290, 8711, 616, 5369, 13, 51403], "temperature": 0.0, "avg_logprob": -0.12729211287064987, "compression_ratio": 1.5324074074074074, "no_speech_prob": 0.0075220344588160515, "words": [{"word": " Maybe", "start": 177.52, "end": 177.82, "probability": 0.9915785789489746}, {"word": " you", "start": 177.82, "end": 178.04, "probability": 0.9878547787666321}, {"word": " could", "start": 178.04, "end": 178.36, "probability": 0.9932275414466858}, {"word": " steal", "start": 178.36, "end": 179.22, "probability": 0.9923547506332397}, {"word": " my", "start": 179.22, "end": 179.9, "probability": 0.9971237778663635}, {"word": " password", "start": 179.9, "end": 180.5, "probability": 0.9993727803230286}, {"word": " or", "start": 180.5, "end": 180.86, "probability": 0.8924471735954285}, {"word": " impersonate", "start": 180.86, "end": 181.42, "probability": 0.9967566728591919}, {"word": " me", "start": 181.42, "end": 181.64, "probability": 0.9976024031639099}, {"word": " and", "start": 181.64, "end": 181.92, "probability": 0.9706672430038452}, {"word": " steal", "start": 181.92, "end": 182.26, "probability": 0.9990209341049194}, {"word": " my", "start": 182.26, "end": 182.64, "probability": 0.9991519451141357}, {"word": " identity.", "start": 182.64, "end": 183.12, "probability": 0.9985376596450806}]}, {"id": 33, "seek": 16230, "start": 183.9, "end": 187.6, "text": " If you share enough data, such things inevitably become possible.", "tokens": [51439, 1002, 345, 2648, 1576, 1366, 11, 884, 1243, 16857, 1716, 1744, 13, 51627], "temperature": 0.0, "avg_logprob": -0.12729211287064987, "compression_ratio": 1.5324074074074074, "no_speech_prob": 0.0075220344588160515, "words": [{"word": " If", "start": 183.9, "end": 184.14, "probability": 0.9971454739570618}, {"word": " you", "start": 184.14, "end": 184.28, "probability": 0.9991004467010498}, {"word": " share", "start": 184.28, "end": 184.46, "probability": 0.9985287189483643}, {"word": " enough", "start": 184.46, "end": 184.68, "probability": 0.9989161491394043}, {"word": " data,", "start": 184.68, "end": 185.04, "probability": 0.9985752105712891}, {"word": " such", "start": 185.4, "end": 185.64, "probability": 0.9734892249107361}, {"word": " things", "start": 185.64, "end": 186.04, "probability": 0.9984725117683411}, {"word": " inevitably", "start": 186.04, "end": 186.4, "probability": 0.9818087220191956}, {"word": " become", "start": 186.4, "end": 187.0, "probability": 0.9970344305038452}, {"word": " possible.", "start": 187.0, "end": 187.6, "probability": 0.9986817240715027}]}, {"id": 34, "seek": 18760, "start": 188.26000000000002, "end": 193.36, "text": " Some might say we shouldn't do away with machine learning and keep everything secret,", "tokens": [50411, 2773, 1244, 910, 356, 6584, 470, 466, 1497, 351, 4572, 4673, 290, 1394, 2279, 3200, 11, 50659], "temperature": 0.0, "avg_logprob": -0.06444989037267941, "compression_ratio": 1.5812274368231047, "no_speech_prob": 0.01897745206952095, "words": [{"word": " Some", "start": 188.26000000000002, "end": 188.86, "probability": 0.6211700439453125}, {"word": " might", "start": 188.86, "end": 189.04, "probability": 0.9696332216262817}, {"word": " say", "start": 189.04, "end": 189.24, "probability": 0.981620728969574}, {"word": " we", "start": 189.24, "end": 189.44, "probability": 0.9174416065216064}, {"word": " shouldn't", "start": 189.44, "end": 189.66, "probability": 0.9348413646221161}, {"word": " do", "start": 189.66, "end": 190.4, "probability": 0.9750003814697266}, {"word": " away", "start": 190.4, "end": 191.24, "probability": 0.8286336660385132}, {"word": " with", "start": 191.24, "end": 191.58, "probability": 0.9929353594779968}, {"word": " machine", "start": 191.58, "end": 191.8, "probability": 0.9044138789176941}, {"word": " learning", "start": 191.8, "end": 192.04, "probability": 0.9877479076385498}, {"word": " and", "start": 192.04, "end": 192.42, "probability": 0.9333767890930176}, {"word": " keep", "start": 192.42, "end": 192.58, "probability": 0.9928029179573059}, {"word": " everything", "start": 192.58, "end": 192.98, "probability": 0.9954581260681152}, {"word": " secret,", "start": 192.98, "end": 193.36, "probability": 0.9735928177833557}]}, {"id": 35, "seek": 18760, "start": 193.7, "end": 196.32, "text": " but machine learning has a great potential for good.", "tokens": [50659, 475, 4572, 4673, 468, 257, 1049, 2785, 329, 922, 13, 50823], "temperature": 0.0, "avg_logprob": -0.06444989037267941, "compression_ratio": 1.5812274368231047, "no_speech_prob": 0.01897745206952095, "words": [{"word": " but", "start": 193.7, "end": 193.84, "probability": 0.9932569265365601}, {"word": " machine", "start": 193.84, "end": 194.08, "probability": 0.9730015397071838}, {"word": " learning", "start": 194.08, "end": 194.36, "probability": 0.9944097399711609}, {"word": " has", "start": 194.36, "end": 194.84, "probability": 0.9892114400863647}, {"word": " a", "start": 194.84, "end": 195.18, "probability": 0.8896937370300293}, {"word": " great", "start": 195.18, "end": 195.48, "probability": 0.9736711382865906}, {"word": " potential", "start": 195.48, "end": 195.84, "probability": 0.9987446069717407}, {"word": " for", "start": 195.84, "end": 196.1, "probability": 0.9953041076660156}, {"word": " good.", "start": 196.1, "end": 196.32, "probability": 0.994541347026825}]}, {"id": 36, "seek": 18760, "start": 197.02, "end": 200.68, "text": " Suppose I find out that I have been growing cancer for the last three months.", "tokens": [50823, 39200, 314, 1064, 503, 326, 314, 423, 587, 3957, 4890, 329, 262, 938, 1115, 1933, 13, 51031], "temperature": 0.0, "avg_logprob": -0.06444989037267941, "compression_ratio": 1.5812274368231047, "no_speech_prob": 0.01897745206952095, "words": [{"word": " Suppose", "start": 197.02, "end": 197.26, "probability": 0.9613398313522339}, {"word": " I", "start": 197.26, "end": 197.52, "probability": 0.9807553887367249}, {"word": " find", "start": 197.52, "end": 197.8, "probability": 0.9514200687408447}, {"word": " out", "start": 197.8, "end": 198.02, "probability": 0.9942007064819336}, {"word": " that", "start": 198.02, "end": 198.24, "probability": 0.9510683417320251}, {"word": " I", "start": 198.24, "end": 198.36, "probability": 0.9947785139083862}, {"word": " have", "start": 198.36, "end": 198.52, "probability": 0.8968207836151123}, {"word": " been", "start": 198.52, "end": 198.68, "probability": 0.9962389469146729}, {"word": " growing", "start": 198.68, "end": 199.16, "probability": 0.9224908947944641}, {"word": " cancer", "start": 199.16, "end": 199.64, "probability": 0.9945279955863953}, {"word": " for", "start": 199.64, "end": 199.9, "probability": 0.9891312122344971}, {"word": " the", "start": 199.9, "end": 200.02, "probability": 0.9947906732559204}, {"word": " last", "start": 200.02, "end": 200.2, "probability": 0.9932221174240112}, {"word": " three", "start": 200.2, "end": 200.42, "probability": 0.897468090057373}, {"word": " months.", "start": 200.42, "end": 200.68, "probability": 0.998196542263031}]}, {"id": 37, "seek": 18760, "start": 201.08, "end": 203.5, "text": " Now my data, including the medical diagnosis,", "tokens": [51031, 2735, 616, 1366, 11, 1390, 262, 3315, 13669, 11, 51171], "temperature": 0.0, "avg_logprob": -0.06444989037267941, "compression_ratio": 1.5812274368231047, "no_speech_prob": 0.01897745206952095, "words": [{"word": " Now", "start": 201.08, "end": 201.28, "probability": 0.9909257888793945}, {"word": " my", "start": 201.28, "end": 201.58, "probability": 0.6750947833061218}, {"word": " data,", "start": 201.58, "end": 201.92, "probability": 0.9989318251609802}, {"word": " including", "start": 202.1, "end": 202.4, "probability": 0.9990774393081665}, {"word": " the", "start": 202.4, "end": 202.62, "probability": 0.9963615536689758}, {"word": " medical", "start": 202.62, "end": 202.92, "probability": 0.9975904226303101}, {"word": " diagnosis,", "start": 202.92, "end": 203.5, "probability": 0.9951103329658508}]}, {"id": 38, "seek": 18760, "start": 203.82, "end": 207.6, "text": " contains knowledge that might help others detect cancer before it is too late.", "tokens": [51171, 4909, 3725, 326, 1244, 1037, 1854, 4886, 4890, 878, 340, 318, 1165, 2739, 13, 51359], "temperature": 0.0, "avg_logprob": -0.06444989037267941, "compression_ratio": 1.5812274368231047, "no_speech_prob": 0.01897745206952095, "words": [{"word": " contains", "start": 203.82, "end": 204.2, "probability": 0.9975638389587402}, {"word": " knowledge", "start": 204.2, "end": 204.66, "probability": 0.9981014132499695}, {"word": " that", "start": 204.66, "end": 204.96, "probability": 0.9986328482627869}, {"word": " might", "start": 204.96, "end": 205.12, "probability": 0.9973512887954712}, {"word": " help", "start": 205.12, "end": 205.4, "probability": 0.9993854761123657}, {"word": " others", "start": 205.4, "end": 205.74, "probability": 0.9946156144142151}, {"word": " detect", "start": 205.74, "end": 206.08, "probability": 0.9839521050453186}, {"word": " cancer", "start": 206.08, "end": 206.6, "probability": 0.9981332421302795}, {"word": " before", "start": 206.6, "end": 206.92, "probability": 0.994228720664978}, {"word": " it", "start": 206.92, "end": 207.02, "probability": 0.9956474900245667}, {"word": " is", "start": 207.02, "end": 207.14, "probability": 0.9793074131011963}, {"word": " too", "start": 207.14, "end": 207.3, "probability": 0.9951196908950806}, {"word": " late.", "start": 207.3, "end": 207.6, "probability": 0.9991637468338013}]}, {"id": 39, "seek": 18760, "start": 208.64, "end": 214.9, "text": " Additionally, as I found out at CVPR this year, the carbon footprint of decentralized frameworks", "tokens": [51403, 12032, 11, 355, 314, 1043, 503, 379, 26196, 4805, 428, 614, 11, 262, 6588, 24713, 286, 26512, 29251, 51727], "temperature": 0.0, "avg_logprob": -0.06444989037267941, "compression_ratio": 1.5812274368231047, "no_speech_prob": 0.01897745206952095, "words": [{"word": " Additionally,", "start": 208.64, "end": 208.94, "probability": 0.9872701168060303}, {"word": " as", "start": 209.0, "end": 209.26, "probability": 0.9962679743766785}, {"word": " I", "start": 209.26, "end": 209.4, "probability": 0.9974926710128784}, {"word": " found", "start": 209.4, "end": 209.58, "probability": 0.9899600744247437}, {"word": " out", "start": 209.58, "end": 209.82, "probability": 0.9959010481834412}, {"word": " at", "start": 209.82, "end": 210.08, "probability": 0.981525182723999}, {"word": " CVPR", "start": 210.08, "end": 210.88, "probability": 0.9884675443172455}, {"word": " this", "start": 210.88, "end": 211.28, "probability": 0.9746142029762268}, {"word": " year,", "start": 211.28, "end": 211.66, "probability": 0.9994070529937744}, {"word": " the", "start": 212.5, "end": 212.8, "probability": 0.9727472066879272}, {"word": " carbon", "start": 212.8, "end": 213.2, "probability": 0.9926750063896179}, {"word": " footprint", "start": 213.2, "end": 213.68, "probability": 0.9994148015975952}, {"word": " of", "start": 213.68, "end": 214.02, "probability": 0.9983976483345032}, {"word": " decentralized", "start": 214.02, "end": 214.44, "probability": 0.9929265379905701}, {"word": " frameworks", "start": 214.44, "end": 214.9, "probability": 0.9950554370880127}]}, {"id": 40, "seek": 21490, "start": 214.9, "end": 222.0, "text": " can actually be lower than centralized systems because more powerful GPUs are less power efficient.", "tokens": [50363, 460, 1682, 307, 2793, 621, 29024, 3341, 780, 517, 3665, 32516, 389, 1342, 1176, 6942, 13, 50719], "temperature": 0.0, "avg_logprob": -0.15690550961337246, "compression_ratio": 1.4618320610687023, "no_speech_prob": 0.017633521929383278, "words": [{"word": " can", "start": 214.9, "end": 215.14, "probability": 0.8111503720283508}, {"word": " actually", "start": 215.14, "end": 215.38, "probability": 0.9546414017677307}, {"word": " be", "start": 215.38, "end": 215.58, "probability": 0.9962812066078186}, {"word": " lower", "start": 215.58, "end": 215.8, "probability": 0.9935478568077087}, {"word": " than", "start": 215.8, "end": 216.08, "probability": 0.9838501214981079}, {"word": " centralized", "start": 216.08, "end": 216.6, "probability": 0.987242579460144}, {"word": " systems", "start": 216.6, "end": 217.94, "probability": 0.8662360906600952}, {"word": " because", "start": 217.94, "end": 218.54, "probability": 0.76761794090271}, {"word": " more", "start": 218.54, "end": 219.06, "probability": 0.9873731732368469}, {"word": " powerful", "start": 219.06, "end": 219.64, "probability": 0.9944709539413452}, {"word": " GPUs", "start": 219.64, "end": 220.06, "probability": 0.9334223866462708}, {"word": " are", "start": 220.06, "end": 220.68, "probability": 0.9195773601531982}, {"word": " less", "start": 220.68, "end": 220.94, "probability": 0.9963370561599731}, {"word": " power", "start": 220.94, "end": 221.52, "probability": 0.9933666586875916}, {"word": " efficient.", "start": 221.52, "end": 222.0, "probability": 0.8135547041893005}]}, {"id": 41, "seek": 21490, "start": 223.34, "end": 228.28, "text": " A lot of their power ends up going into active cooling.", "tokens": [50783, 317, 1256, 286, 511, 1176, 5645, 510, 1016, 656, 4075, 15134, 13, 51031], "temperature": 0.0, "avg_logprob": -0.15690550961337246, "compression_ratio": 1.4618320610687023, "no_speech_prob": 0.017633521929383278, "words": [{"word": " A", "start": 223.34, "end": 223.54, "probability": 0.9487414956092834}, {"word": " lot", "start": 223.54, "end": 223.76, "probability": 0.9993457198143005}, {"word": " of", "start": 223.76, "end": 223.94, "probability": 0.9995631575584412}, {"word": " their", "start": 223.94, "end": 224.06, "probability": 0.9963778853416443}, {"word": " power", "start": 224.06, "end": 224.38, "probability": 0.9987718462944031}, {"word": " ends", "start": 224.38, "end": 224.72, "probability": 0.9965169429779053}, {"word": " up", "start": 224.72, "end": 225.02, "probability": 0.9992788434028625}, {"word": " going", "start": 225.02, "end": 225.48, "probability": 0.9985440969467163}, {"word": " into", "start": 225.48, "end": 226.38, "probability": 0.9915140867233276}, {"word": " active", "start": 226.38, "end": 227.72, "probability": 0.9767370820045471}, {"word": " cooling.", "start": 227.72, "end": 228.28, "probability": 0.996113657951355}]}, {"id": 42, "seek": 21490, "start": 229.2, "end": 235.22, "text": " This is a plot that was provided by Nicholas Lane from the University of Cambridge and Samsung at AI.", "tokens": [51083, 770, 318, 257, 7110, 326, 373, 2810, 416, 20320, 15016, 422, 262, 2059, 286, 14457, 290, 10397, 379, 9552, 13, 51375], "temperature": 0.0, "avg_logprob": -0.15690550961337246, "compression_ratio": 1.4618320610687023, "no_speech_prob": 0.017633521929383278, "words": [{"word": " This", "start": 229.2, "end": 229.58, "probability": 0.9890785813331604}, {"word": " is", "start": 229.58, "end": 229.78, "probability": 0.9977789521217346}, {"word": " a", "start": 229.78, "end": 230.0, "probability": 0.9991267323493958}, {"word": " plot", "start": 230.0, "end": 230.4, "probability": 0.9965237975120544}, {"word": " that", "start": 230.4, "end": 231.08, "probability": 0.9930481910705566}, {"word": " was", "start": 231.08, "end": 231.28, "probability": 0.9991472959518433}, {"word": " provided", "start": 231.28, "end": 231.6, "probability": 0.9976068735122681}, {"word": " by", "start": 231.6, "end": 231.86, "probability": 0.9980789422988892}, {"word": " Nicholas", "start": 231.86, "end": 232.24, "probability": 0.8708387017250061}, {"word": " Lane", "start": 232.24, "end": 232.6, "probability": 0.9360295534133911}, {"word": " from", "start": 232.6, "end": 232.88, "probability": 0.9784883260726929}, {"word": " the", "start": 232.88, "end": 232.98, "probability": 0.9787976145744324}, {"word": " University", "start": 232.98, "end": 233.32, "probability": 0.9805380702018738}, {"word": " of", "start": 233.32, "end": 233.6, "probability": 0.9978412389755249}, {"word": " Cambridge", "start": 233.6, "end": 234.0, "probability": 0.9995577931404114}, {"word": " and", "start": 234.0, "end": 234.28, "probability": 0.5660107135772705}, {"word": " Samsung", "start": 234.28, "end": 234.72, "probability": 0.9198049306869507}, {"word": " at", "start": 234.72, "end": 234.92, "probability": 0.7642095685005188}, {"word": " AI.", "start": 234.92, "end": 235.22, "probability": 0.9605572819709778}]}, {"id": 43, "seek": 21490, "start": 235.96, "end": 242.08, "text": " He presented this at CVPR 2023, less than a week ago, in the Federated Learning Workshop.", "tokens": [51443, 679, 5545, 428, 379, 26196, 4805, 1160, 1954, 11, 1342, 621, 257, 1285, 2084, 11, 287, 262, 35089, 515, 18252, 26701, 13, 51719], "temperature": 0.0, "avg_logprob": -0.15690550961337246, "compression_ratio": 1.4618320610687023, "no_speech_prob": 0.017633521929383278, "words": [{"word": " He", "start": 235.96, "end": 236.12, "probability": 0.39998453855514526}, {"word": " presented", "start": 236.12, "end": 237.02, "probability": 0.9912439584732056}, {"word": " this", "start": 237.02, "end": 237.3, "probability": 0.9874873161315918}, {"word": " at", "start": 237.3, "end": 237.5, "probability": 0.7925485968589783}, {"word": " CVPR", "start": 237.5, "end": 238.14, "probability": 0.9526187181472778}, {"word": " 2023,", "start": 238.14, "end": 238.94, "probability": 0.8840397298336029}, {"word": " less", "start": 239.26, "end": 239.48, "probability": 0.9933742880821228}, {"word": " than", "start": 239.48, "end": 239.7, "probability": 0.9990350008010864}, {"word": " a", "start": 239.7, "end": 239.78, "probability": 0.9988738894462585}, {"word": " week", "start": 239.78, "end": 239.88, "probability": 0.999334990978241}, {"word": " ago,", "start": 239.88, "end": 240.28, "probability": 0.9988401532173157}, {"word": " in", "start": 240.28, "end": 240.78, "probability": 0.9786763787269592}, {"word": " the", "start": 240.78, "end": 241.0, "probability": 0.9914339780807495}, {"word": " Federated", "start": 241.0, "end": 241.4, "probability": 0.9151340425014496}, {"word": " Learning", "start": 241.4, "end": 241.58, "probability": 0.9894371032714844}, {"word": " Workshop.", "start": 241.58, "end": 242.08, "probability": 0.9701253771781921}]}, {"id": 44, "seek": 21490, "start": 242.34, "end": 243.42, "text": " Can you say the name one more time?", "tokens": [51719, 1680, 345, 910, 262, 1438, 530, 517, 640, 30, 51807], "temperature": 0.0, "avg_logprob": -0.15690550961337246, "compression_ratio": 1.4618320610687023, "no_speech_prob": 0.017633521929383278, "words": [{"word": " Can", "start": 242.34, "end": 242.42, "probability": 0.3209560513496399}, {"word": " you", "start": 242.42, "end": 242.52, "probability": 0.9820573329925537}, {"word": " say", "start": 242.52, "end": 242.64, "probability": 0.9944193959236145}, {"word": " the", "start": 242.64, "end": 242.78, "probability": 0.9882901906967163}, {"word": " name", "start": 242.78, "end": 242.9, "probability": 0.9956022500991821}, {"word": " one", "start": 242.9, "end": 243.1, "probability": 0.8200472593307495}, {"word": " more", "start": 243.1, "end": 243.26, "probability": 0.9854307174682617}, {"word": " time?", "start": 243.26, "end": 243.42, "probability": 0.7125886082649231}]}, {"id": 45, "seek": 24342, "start": 243.42, "end": 245.0, "text": " The name of the person?", "tokens": [50387, 383, 1438, 286, 262, 1048, 30, 50439], "temperature": 0.0, "avg_logprob": -0.19219270017411974, "compression_ratio": 1.5852534562211982, "no_speech_prob": 0.015911299735307693, "words": [{"word": " The", "start": 243.42, "end": 244.16, "probability": 0.047099366784095764}, {"word": " name", "start": 244.16, "end": 244.44, "probability": 0.932488739490509}, {"word": " of", "start": 244.44, "end": 244.56, "probability": 0.9919018149375916}, {"word": " the", "start": 244.56, "end": 244.66, "probability": 0.9945899248123169}, {"word": " person?", "start": 244.66, "end": 245.0, "probability": 0.9897215962409973}]}, {"id": 46, "seek": 24342, "start": 245.9, "end": 246.96, "text": " Nicholas Lane.", "tokens": [50499, 20320, 15016, 13, 50539], "temperature": 0.0, "avg_logprob": -0.19219270017411974, "compression_ratio": 1.5852534562211982, "no_speech_prob": 0.015911299735307693, "words": [{"word": " Nicholas", "start": 245.9, "end": 246.5, "probability": 0.41124898195266724}, {"word": " Lane.", "start": 246.5, "end": 246.96, "probability": 0.8955788612365723}]}, {"id": 47, "seek": 24342, "start": 252.06, "end": 258.42, "text": " So, Federated Learning is a machine learning framework which allows you to train neural networks without centralizing data.", "tokens": [50795, 1406, 11, 35089, 515, 18252, 318, 257, 4572, 4673, 9355, 543, 3578, 345, 284, 4512, 17019, 7686, 1231, 4318, 2890, 1366, 13, 51139], "temperature": 0.0, "avg_logprob": -0.19219270017411974, "compression_ratio": 1.5852534562211982, "no_speech_prob": 0.015911299735307693, "words": [{"word": " So,", "start": 251.92000000000002, "end": 252.3, "probability": 0.2370155155658722}, {"word": " Federated", "start": 252.68, "end": 253.0, "probability": 0.4984041934112611}, {"word": " Learning", "start": 253.0, "end": 253.24, "probability": 0.9072222709655762}, {"word": " is", "start": 253.24, "end": 253.74, "probability": 0.9644594788551331}, {"word": " a", "start": 253.74, "end": 253.86, "probability": 0.9854996204376221}, {"word": " machine", "start": 253.86, "end": 254.12, "probability": 0.9192994236946106}, {"word": " learning", "start": 254.12, "end": 254.42, "probability": 0.9358095526695251}, {"word": " framework", "start": 254.42, "end": 254.82, "probability": 0.9987764954566956}, {"word": " which", "start": 254.82, "end": 255.14, "probability": 0.8402607440948486}, {"word": " allows", "start": 255.14, "end": 255.4, "probability": 0.9968786239624023}, {"word": " you", "start": 255.4, "end": 255.6, "probability": 0.996809184551239}, {"word": " to", "start": 255.6, "end": 255.76, "probability": 0.9972298741340637}, {"word": " train", "start": 255.76, "end": 256.16, "probability": 0.9900744557380676}, {"word": " neural", "start": 256.16, "end": 256.56, "probability": 0.9924355149269104}, {"word": " networks", "start": 256.56, "end": 256.94, "probability": 0.9957160353660583}, {"word": " without", "start": 256.94, "end": 257.44, "probability": 0.9908068776130676}, {"word": " centralizing", "start": 257.44, "end": 258.08, "probability": 0.9768210351467133}, {"word": " data.", "start": 258.08, "end": 258.42, "probability": 0.9960839748382568}]}, {"id": 48, "seek": 24342, "start": 258.94, "end": 263.48, "text": " It actually achieves the same performance as centralized machine learning methods.", "tokens": [51139, 632, 1682, 41885, 262, 976, 2854, 355, 29024, 4572, 4673, 5050, 13, 51383], "temperature": 0.0, "avg_logprob": -0.19219270017411974, "compression_ratio": 1.5852534562211982, "no_speech_prob": 0.015911299735307693, "words": [{"word": " It", "start": 258.94, "end": 259.12, "probability": 0.9912728667259216}, {"word": " actually", "start": 259.12, "end": 259.56, "probability": 0.986574113368988}, {"word": " achieves", "start": 259.56, "end": 260.36, "probability": 0.9972242116928101}, {"word": " the", "start": 260.36, "end": 260.56, "probability": 0.9987144470214844}, {"word": " same", "start": 260.56, "end": 260.84, "probability": 0.9990319013595581}, {"word": " performance", "start": 260.84, "end": 261.32, "probability": 0.9954413175582886}, {"word": " as", "start": 261.32, "end": 261.72, "probability": 0.9868740439414978}, {"word": " centralized", "start": 261.72, "end": 262.24, "probability": 0.9917556047439575}, {"word": " machine", "start": 262.24, "end": 262.78, "probability": 0.9742076396942139}, {"word": " learning", "start": 262.78, "end": 263.04, "probability": 0.9877820611000061}, {"word": " methods.", "start": 263.04, "end": 263.48, "probability": 0.9953227639198303}]}, {"id": 49, "seek": 24342, "start": 263.88, "end": 269.36, "text": " It was first designed to contend with laws that prohibit the unauthorized sharing of medical data.", "tokens": [51383, 632, 373, 717, 3562, 284, 24351, 351, 3657, 326, 20767, 262, 22959, 7373, 286, 3315, 1366, 13, 51659], "temperature": 0.0, "avg_logprob": -0.19219270017411974, "compression_ratio": 1.5852534562211982, "no_speech_prob": 0.015911299735307693, "words": [{"word": " It", "start": 263.88, "end": 264.0, "probability": 0.990770161151886}, {"word": " was", "start": 264.0, "end": 264.1, "probability": 0.9980973601341248}, {"word": " first", "start": 264.1, "end": 264.34, "probability": 0.9978212118148804}, {"word": " designed", "start": 264.34, "end": 264.86, "probability": 0.9966732263565063}, {"word": " to", "start": 264.86, "end": 265.14, "probability": 0.9983069896697998}, {"word": " contend", "start": 265.14, "end": 265.54, "probability": 0.9974697828292847}, {"word": " with", "start": 265.54, "end": 265.9, "probability": 0.9990146160125732}, {"word": " laws", "start": 265.9, "end": 266.34, "probability": 0.9921485185623169}, {"word": " that", "start": 266.34, "end": 266.6, "probability": 0.9954807758331299}, {"word": " prohibit", "start": 266.6, "end": 266.94, "probability": 0.9726430773735046}, {"word": " the", "start": 266.94, "end": 267.36, "probability": 0.9840108752250671}, {"word": " unauthorized", "start": 267.36, "end": 267.9, "probability": 0.9758836627006531}, {"word": " sharing", "start": 267.9, "end": 268.4, "probability": 0.9979178309440613}, {"word": " of", "start": 268.4, "end": 268.6, "probability": 0.9990707635879517}, {"word": " medical", "start": 268.6, "end": 268.9, "probability": 0.997489333152771}, {"word": " data.", "start": 268.9, "end": 269.36, "probability": 0.9986279010772705}]}, {"id": 50, "seek": 26936, "start": 270.08000000000004, "end": 275.96, "text": " Hospitals trained neural networks locally and communicated those parameters to a central server.", "tokens": [50399, 39946, 8321, 8776, 17019, 7686, 15726, 290, 28412, 883, 10007, 284, 257, 4318, 4382, 13, 50715], "temperature": 0.0, "avg_logprob": -0.1011196772257487, "compression_ratio": 1.7377777777777779, "no_speech_prob": 0.00241907499730587, "words": [{"word": " Hospitals", "start": 270.08000000000004, "end": 270.66, "probability": 0.8336373567581177}, {"word": " trained", "start": 270.66, "end": 271.04, "probability": 0.49376365542411804}, {"word": " neural", "start": 271.04, "end": 271.32, "probability": 0.9870217442512512}, {"word": " networks", "start": 271.32, "end": 271.76, "probability": 0.996076762676239}, {"word": " locally", "start": 271.76, "end": 272.28, "probability": 0.9728366732597351}, {"word": " and", "start": 272.28, "end": 273.26, "probability": 0.5780479907989502}, {"word": " communicated", "start": 273.26, "end": 273.84, "probability": 0.9872318506240845}, {"word": " those", "start": 273.84, "end": 274.12, "probability": 0.9756929278373718}, {"word": " parameters", "start": 274.12, "end": 274.62, "probability": 0.9973178505897522}, {"word": " to", "start": 274.62, "end": 275.12, "probability": 0.9930974841117859}, {"word": " a", "start": 275.12, "end": 275.26, "probability": 0.9923744797706604}, {"word": " central", "start": 275.26, "end": 275.54, "probability": 0.974980354309082}, {"word": " server.", "start": 275.54, "end": 275.96, "probability": 0.9981254935264587}]}, {"id": 51, "seek": 26936, "start": 276.48, "end": 280.4, "text": " The central server then combined and redistributed the resulting parameters.", "tokens": [50715, 383, 4318, 4382, 788, 5929, 290, 38913, 262, 7186, 10007, 13, 50931], "temperature": 0.0, "avg_logprob": -0.1011196772257487, "compression_ratio": 1.7377777777777779, "no_speech_prob": 0.00241907499730587, "words": [{"word": " The", "start": 276.48, "end": 276.64, "probability": 0.9913416504859924}, {"word": " central", "start": 276.64, "end": 276.9, "probability": 0.9937291145324707}, {"word": " server", "start": 276.9, "end": 277.28, "probability": 0.9977009892463684}, {"word": " then", "start": 277.28, "end": 277.52, "probability": 0.954088032245636}, {"word": " combined", "start": 277.52, "end": 278.06, "probability": 0.9838942885398865}, {"word": " and", "start": 278.06, "end": 278.36, "probability": 0.9917297959327698}, {"word": " redistributed", "start": 278.36, "end": 278.94, "probability": 0.9833773374557495}, {"word": " the", "start": 278.94, "end": 279.28, "probability": 0.9958614706993103}, {"word": " resulting", "start": 279.28, "end": 279.78, "probability": 0.9976192116737366}, {"word": " parameters.", "start": 279.78, "end": 280.4, "probability": 0.9965693950653076}]}, {"id": 52, "seek": 26936, "start": 280.86, "end": 285.9, "text": " This process is repeated until the combined parameters stop improving.", "tokens": [50931, 770, 1429, 318, 5100, 1566, 262, 5929, 10007, 2245, 10068, 13, 51191], "temperature": 0.0, "avg_logprob": -0.1011196772257487, "compression_ratio": 1.7377777777777779, "no_speech_prob": 0.00241907499730587, "words": [{"word": " This", "start": 280.86, "end": 281.02, "probability": 0.9941674470901489}, {"word": " process", "start": 281.02, "end": 281.44, "probability": 0.998175859451294}, {"word": " is", "start": 281.44, "end": 281.7, "probability": 0.9976872205734253}, {"word": " repeated", "start": 281.7, "end": 282.1, "probability": 0.9992973804473877}, {"word": " until", "start": 282.1, "end": 282.58, "probability": 0.996199905872345}, {"word": " the", "start": 282.58, "end": 282.72, "probability": 0.9831499457359314}, {"word": " combined", "start": 282.72, "end": 283.34, "probability": 0.9984456896781921}, {"word": " parameters", "start": 283.34, "end": 284.92, "probability": 0.908142626285553}, {"word": " stop", "start": 284.92, "end": 285.34, "probability": 0.9579210877418518}, {"word": " improving.", "start": 285.34, "end": 285.9, "probability": 0.9967908263206482}]}, {"id": 53, "seek": 26936, "start": 286.62, "end": 287.58, "text": " Here we have a diagram.", "tokens": [51219, 3423, 356, 423, 257, 16362, 13, 51275], "temperature": 0.0, "avg_logprob": -0.1011196772257487, "compression_ratio": 1.7377777777777779, "no_speech_prob": 0.00241907499730587, "words": [{"word": " Here", "start": 286.62, "end": 286.8, "probability": 0.9936596751213074}, {"word": " we", "start": 286.8, "end": 286.9, "probability": 0.8964202404022217}, {"word": " have", "start": 286.9, "end": 287.06, "probability": 0.9993182420730591}, {"word": " a", "start": 287.06, "end": 287.18, "probability": 0.9922410249710083}, {"word": " diagram.", "start": 287.18, "end": 287.58, "probability": 0.9943957328796387}]}, {"id": 54, "seek": 26936, "start": 289.0, "end": 294.62, "text": " Notice that each line here is a hospital and that's the training happening.", "tokens": [51327, 17641, 326, 1123, 1627, 994, 318, 257, 4436, 290, 326, 338, 262, 3047, 5836, 13, 51627], "temperature": 0.0, "avg_logprob": -0.1011196772257487, "compression_ratio": 1.7377777777777779, "no_speech_prob": 0.00241907499730587, "words": [{"word": " Notice", "start": 289.0, "end": 289.0, "probability": 0.6708493232727051}, {"word": " that", "start": 289.0, "end": 289.5, "probability": 0.9629799723625183}, {"word": " each", "start": 289.5, "end": 290.16, "probability": 0.3549608588218689}, {"word": " line", "start": 290.16, "end": 290.5, "probability": 0.9967356324195862}, {"word": " here", "start": 290.5, "end": 290.84, "probability": 0.9751244187355042}, {"word": " is", "start": 290.84, "end": 291.34, "probability": 0.9887257218360901}, {"word": " a", "start": 291.34, "end": 291.54, "probability": 0.9911332726478577}, {"word": " hospital", "start": 291.54, "end": 291.98, "probability": 0.9908513426780701}, {"word": " and", "start": 291.98, "end": 292.9, "probability": 0.6214175224304199}, {"word": " that's", "start": 292.9, "end": 293.5, "probability": 0.9264482259750366}, {"word": " the", "start": 293.5, "end": 293.82, "probability": 0.9982579350471497}, {"word": " training", "start": 293.82, "end": 294.2, "probability": 0.9982352256774902}, {"word": " happening.", "start": 294.2, "end": 294.62, "probability": 0.9974321722984314}]}, {"id": 55, "seek": 26936, "start": 294.84, "end": 296.82, "text": " So you're converting the data into parameters.", "tokens": [51627, 1406, 345, 821, 23202, 262, 1366, 656, 10007, 13, 51747], "temperature": 0.0, "avg_logprob": -0.1011196772257487, "compression_ratio": 1.7377777777777779, "no_speech_prob": 0.00241907499730587, "words": [{"word": " So", "start": 294.84, "end": 294.86, "probability": 0.7954946756362915}, {"word": " you're", "start": 294.86, "end": 295.02, "probability": 0.8951645791530609}, {"word": " converting", "start": 295.02, "end": 295.36, "probability": 0.99747234582901}, {"word": " the", "start": 295.36, "end": 295.56, "probability": 0.9969172477722168}, {"word": " data", "start": 295.56, "end": 295.82, "probability": 0.9995336532592773}, {"word": " into", "start": 295.82, "end": 296.3, "probability": 0.9973770380020142}, {"word": " parameters.", "start": 296.3, "end": 296.82, "probability": 0.9947962164878845}]}, {"id": 56, "seek": 29682, "start": 296.82, "end": 304.62, "text": " The data never leaves the hospital and it is eventually aggregated in a central server.", "tokens": [50375, 383, 1366, 1239, 5667, 262, 4436, 290, 340, 318, 4191, 13262, 515, 287, 257, 4318, 4382, 13, 50751], "temperature": 0.0, "avg_logprob": -0.12298134753578588, "compression_ratio": 1.7389162561576355, "no_speech_prob": 0.006620736327022314, "words": [{"word": " The", "start": 296.82, "end": 297.28, "probability": 0.3913857936859131}, {"word": " data", "start": 297.28, "end": 297.5, "probability": 0.9691938161849976}, {"word": " never", "start": 297.5, "end": 297.9, "probability": 0.9934008717536926}, {"word": " leaves", "start": 297.9, "end": 298.34, "probability": 0.997087299823761}, {"word": " the", "start": 298.34, "end": 298.54, "probability": 0.9984869956970215}, {"word": " hospital", "start": 298.54, "end": 299.12, "probability": 0.9984028935432434}, {"word": " and", "start": 299.12, "end": 300.4, "probability": 0.5517793893814087}, {"word": " it", "start": 300.4, "end": 300.66, "probability": 0.9342145919799805}, {"word": " is", "start": 300.66, "end": 300.92, "probability": 0.9824715256690979}, {"word": " eventually", "start": 300.92, "end": 301.42, "probability": 0.8726226687431335}, {"word": " aggregated", "start": 301.42, "end": 303.68, "probability": 0.7495451271533966}, {"word": " in", "start": 303.68, "end": 303.86, "probability": 0.9828091263771057}, {"word": " a", "start": 303.86, "end": 303.98, "probability": 0.986669659614563}, {"word": " central", "start": 303.98, "end": 304.2, "probability": 0.9783631563186646}, {"word": " server.", "start": 304.2, "end": 304.62, "probability": 0.9984478950500488}]}, {"id": 57, "seek": 29682, "start": 306.64, "end": 312.68, "text": " The downside here is that it does still require a central server.", "tokens": [50875, 383, 25320, 994, 318, 326, 340, 857, 991, 2421, 257, 4318, 4382, 13, 51155], "temperature": 0.0, "avg_logprob": -0.12298134753578588, "compression_ratio": 1.7389162561576355, "no_speech_prob": 0.006620736327022314, "words": [{"word": " The", "start": 306.64, "end": 307.34, "probability": 0.9904952049255371}, {"word": " downside", "start": 307.34, "end": 307.7, "probability": 0.9638614654541016}, {"word": " here", "start": 307.7, "end": 308.3, "probability": 0.9840084314346313}, {"word": " is", "start": 308.3, "end": 308.64, "probability": 0.9849733114242554}, {"word": " that", "start": 308.64, "end": 308.9, "probability": 0.9675313234329224}, {"word": " it", "start": 308.9, "end": 310.68, "probability": 0.9757290482521057}, {"word": " does", "start": 310.68, "end": 310.98, "probability": 0.7351875901222229}, {"word": " still", "start": 310.98, "end": 311.44, "probability": 0.9901061654090881}, {"word": " require", "start": 311.44, "end": 311.8, "probability": 0.996982753276825}, {"word": " a", "start": 311.8, "end": 312.06, "probability": 0.9943431615829468}, {"word": " central", "start": 312.06, "end": 312.3, "probability": 0.9974532723426819}, {"word": " server.", "start": 312.3, "end": 312.68, "probability": 0.9986903071403503}]}, {"id": 58, "seek": 29682, "start": 314.44, "end": 320.74, "text": " The basic idea of decentralized federated learning is that instead of exchanging parameters with a central server,", "tokens": [51239, 383, 4096, 2126, 286, 26512, 28062, 515, 4673, 318, 326, 2427, 286, 39167, 10007, 351, 257, 4318, 4382, 11, 51571], "temperature": 0.0, "avg_logprob": -0.12298134753578588, "compression_ratio": 1.7389162561576355, "no_speech_prob": 0.006620736327022314, "words": [{"word": " The", "start": 314.44, "end": 314.64, "probability": 0.995134174823761}, {"word": " basic", "start": 314.64, "end": 314.9, "probability": 0.9955489635467529}, {"word": " idea", "start": 314.9, "end": 315.32, "probability": 0.9952124357223511}, {"word": " of", "start": 315.32, "end": 315.78, "probability": 0.6895835399627686}, {"word": " decentralized", "start": 315.78, "end": 316.16, "probability": 0.7392138838768005}, {"word": " federated", "start": 316.16, "end": 316.9, "probability": 0.9698076248168945}, {"word": " learning", "start": 316.9, "end": 317.32, "probability": 0.9979621171951294}, {"word": " is", "start": 317.32, "end": 317.8, "probability": 0.9838362336158752}, {"word": " that", "start": 317.8, "end": 317.98, "probability": 0.9904652833938599}, {"word": " instead", "start": 317.98, "end": 318.4, "probability": 0.951705276966095}, {"word": " of", "start": 318.4, "end": 318.64, "probability": 0.9983648657798767}, {"word": " exchanging", "start": 318.64, "end": 319.12, "probability": 0.9962947964668274}, {"word": " parameters", "start": 319.12, "end": 319.62, "probability": 0.9831474423408508}, {"word": " with", "start": 319.62, "end": 319.84, "probability": 0.9976981282234192}, {"word": " a", "start": 319.84, "end": 320.0, "probability": 0.6973006725311279}, {"word": " central", "start": 320.0, "end": 320.28, "probability": 0.9983589053153992}, {"word": " server,", "start": 320.28, "end": 320.74, "probability": 0.9991028308868408}]}, {"id": 59, "seek": 29682, "start": 321.14, "end": 325.52, "text": " you treat each participant as a node in a graph and share with your graph neighbors.", "tokens": [51571, 345, 2190, 1123, 18399, 355, 257, 10139, 287, 257, 4823, 290, 2648, 351, 534, 4823, 12020, 13, 51803], "temperature": 0.0, "avg_logprob": -0.12298134753578588, "compression_ratio": 1.7389162561576355, "no_speech_prob": 0.006620736327022314, "words": [{"word": " you", "start": 321.14, "end": 321.24, "probability": 0.9921060800552368}, {"word": " treat", "start": 321.24, "end": 321.44, "probability": 0.9905755519866943}, {"word": " each", "start": 321.44, "end": 321.72, "probability": 0.999243974685669}, {"word": " participant", "start": 321.72, "end": 322.32, "probability": 0.9985798597335815}, {"word": " as", "start": 322.32, "end": 323.04, "probability": 0.990913987159729}, {"word": " a", "start": 323.04, "end": 323.16, "probability": 0.9980539083480835}, {"word": " node", "start": 323.16, "end": 323.38, "probability": 0.9940312504768372}, {"word": " in", "start": 323.38, "end": 323.56, "probability": 0.9042508602142334}, {"word": " a", "start": 323.56, "end": 323.7, "probability": 0.9956809282302856}, {"word": " graph", "start": 323.7, "end": 323.9, "probability": 0.9962242841720581}, {"word": " and", "start": 323.9, "end": 324.26, "probability": 0.9164955019950867}, {"word": " share", "start": 324.26, "end": 324.54, "probability": 0.9990432858467102}, {"word": " with", "start": 324.54, "end": 324.74, "probability": 0.9949178695678711}, {"word": " your", "start": 324.74, "end": 324.92, "probability": 0.9982399940490723}, {"word": " graph", "start": 324.92, "end": 325.18, "probability": 0.9805025458335876}, {"word": " neighbors.", "start": 325.18, "end": 325.52, "probability": 0.976565957069397}]}, {"id": 60, "seek": 32552, "start": 325.52, "end": 327.62, "text": " There are many ways to do this.", "tokens": [50407, 1318, 389, 867, 2842, 284, 466, 428, 13, 50475], "temperature": 0.0, "avg_logprob": -0.09193602204322815, "compression_ratio": 1.5720524017467248, "no_speech_prob": 0.0036252900026738644, "words": [{"word": " There", "start": 325.52, "end": 326.24, "probability": 0.45680946111679077}, {"word": " are", "start": 326.24, "end": 326.78, "probability": 0.9859848022460938}, {"word": " many", "start": 326.78, "end": 326.88, "probability": 0.9904148578643799}, {"word": " ways", "start": 326.88, "end": 327.08, "probability": 0.9955728054046631}, {"word": " to", "start": 327.08, "end": 327.26, "probability": 0.9940286874771118}, {"word": " do", "start": 327.26, "end": 327.42, "probability": 0.9923191666603088}, {"word": " this.", "start": 327.42, "end": 327.62, "probability": 0.9904499650001526}]}, {"id": 61, "seek": 32552, "start": 328.04, "end": 328.92, "text": " Some suggest blockchain.", "tokens": [50475, 2773, 1950, 11779, 13, 50543], "temperature": 0.0, "avg_logprob": -0.09193602204322815, "compression_ratio": 1.5720524017467248, "no_speech_prob": 0.0036252900026738644, "words": [{"word": " Some", "start": 328.04, "end": 328.06, "probability": 0.965586245059967}, {"word": " suggest", "start": 328.06, "end": 328.44, "probability": 0.9762312769889832}, {"word": " blockchain.", "start": 328.44, "end": 328.92, "probability": 0.9364418983459473}]}, {"id": 62, "seek": 32552, "start": 329.36, "end": 330.4, "text": " I think blockchain is overkill.", "tokens": [50543, 314, 892, 11779, 318, 625, 12728, 13, 50627], "temperature": 0.0, "avg_logprob": -0.09193602204322815, "compression_ratio": 1.5720524017467248, "no_speech_prob": 0.0036252900026738644, "words": [{"word": " I", "start": 329.36, "end": 329.36, "probability": 0.9759806394577026}, {"word": " think", "start": 329.36, "end": 329.54, "probability": 0.9973670840263367}, {"word": " blockchain", "start": 329.54, "end": 329.9, "probability": 0.9783194661140442}, {"word": " is", "start": 329.9, "end": 330.06, "probability": 0.9935018420219421}, {"word": " overkill.", "start": 330.06, "end": 330.4, "probability": 0.9820828437805176}]}, {"id": 63, "seek": 32552, "start": 331.02, "end": 332.12, "text": " Maybe Holochain is the answer.", "tokens": [50627, 6674, 6479, 5374, 391, 318, 262, 3280, 13, 50691], "temperature": 0.0, "avg_logprob": -0.09193602204322815, "compression_ratio": 1.5720524017467248, "no_speech_prob": 0.0036252900026738644, "words": [{"word": " Maybe", "start": 331.02, "end": 331.14, "probability": 0.9832467436790466}, {"word": " Holochain", "start": 331.14, "end": 331.58, "probability": 0.8968265056610107}, {"word": " is", "start": 331.58, "end": 331.74, "probability": 0.9850220084190369}, {"word": " the", "start": 331.74, "end": 331.86, "probability": 0.9979501366615295}, {"word": " answer.", "start": 331.86, "end": 332.12, "probability": 0.9986158609390259}]}, {"id": 64, "seek": 32552, "start": 333.8, "end": 335.12, "text": " Here I have an example graph.", "tokens": [50779, 3423, 314, 423, 281, 1672, 4823, 13, 50843], "temperature": 0.0, "avg_logprob": -0.09193602204322815, "compression_ratio": 1.5720524017467248, "no_speech_prob": 0.0036252900026738644, "words": [{"word": " Here", "start": 333.8, "end": 334.12, "probability": 0.9852729439735413}, {"word": " I", "start": 334.12, "end": 334.26, "probability": 0.7885800004005432}, {"word": " have", "start": 334.26, "end": 334.4, "probability": 0.9961686730384827}, {"word": " an", "start": 334.4, "end": 334.48, "probability": 0.9973021745681763}, {"word": " example", "start": 334.48, "end": 334.8, "probability": 0.998816728591919}, {"word": " graph.", "start": 334.8, "end": 335.12, "probability": 0.9964160919189453}]}, {"id": 65, "seek": 32552, "start": 335.94, "end": 339.92, "text": " In this case, instead of hospitals, I want you to think of each node as a person.", "tokens": [50875, 554, 428, 1339, 11, 2427, 286, 11301, 11, 314, 765, 345, 284, 892, 286, 1123, 10139, 355, 257, 1048, 13, 51083], "temperature": 0.0, "avg_logprob": -0.09193602204322815, "compression_ratio": 1.5720524017467248, "no_speech_prob": 0.0036252900026738644, "words": [{"word": " In", "start": 335.94, "end": 336.0, "probability": 0.970339834690094}, {"word": " this", "start": 336.0, "end": 336.18, "probability": 0.998934805393219}, {"word": " case,", "start": 336.18, "end": 336.46, "probability": 0.9994792342185974}, {"word": " instead", "start": 336.9, "end": 337.18, "probability": 0.9941978454589844}, {"word": " of", "start": 337.18, "end": 337.44, "probability": 0.9994194507598877}, {"word": " hospitals,", "start": 337.44, "end": 337.96, "probability": 0.9553104639053345}, {"word": " I", "start": 338.24, "end": 338.26, "probability": 0.9970211386680603}, {"word": " want", "start": 338.26, "end": 338.38, "probability": 0.9981746673583984}, {"word": " you", "start": 338.38, "end": 338.5, "probability": 0.9961776733398438}, {"word": " to", "start": 338.5, "end": 338.62, "probability": 0.9942266941070557}, {"word": " think", "start": 338.62, "end": 338.72, "probability": 0.9988018274307251}, {"word": " of", "start": 338.72, "end": 338.88, "probability": 0.9931178092956543}, {"word": " each", "start": 338.88, "end": 339.04, "probability": 0.9945735335350037}, {"word": " node", "start": 339.04, "end": 339.28, "probability": 0.9233330488204956}, {"word": " as", "start": 339.28, "end": 339.52, "probability": 0.9924760460853577}, {"word": " a", "start": 339.52, "end": 339.64, "probability": 0.9983785152435303}, {"word": " person.", "start": 339.64, "end": 339.92, "probability": 0.9973003268241882}]}, {"id": 66, "seek": 32552, "start": 340.68, "end": 345.04, "text": " Each node shares knowledge with adjacent nodes and has a certain degree of trust", "tokens": [51115, 5501, 10139, 7303, 3725, 351, 15909, 13760, 290, 468, 257, 1728, 4922, 286, 3774, 51335], "temperature": 0.0, "avg_logprob": -0.09193602204322815, "compression_ratio": 1.5720524017467248, "no_speech_prob": 0.0036252900026738644, "words": [{"word": " Each", "start": 340.68, "end": 340.88, "probability": 0.9979972243309021}, {"word": " node", "start": 340.88, "end": 341.22, "probability": 0.9954158067703247}, {"word": " shares", "start": 341.22, "end": 341.58, "probability": 0.9934095144271851}, {"word": " knowledge", "start": 341.58, "end": 342.0, "probability": 0.9968626499176025}, {"word": " with", "start": 342.0, "end": 342.3, "probability": 0.996156632900238}, {"word": " adjacent", "start": 342.3, "end": 342.58, "probability": 0.9851436018943787}, {"word": " nodes", "start": 342.58, "end": 343.06, "probability": 0.9978384375572205}, {"word": " and", "start": 343.06, "end": 343.56, "probability": 0.7790934443473816}, {"word": " has", "start": 343.56, "end": 343.82, "probability": 0.9974666833877563}, {"word": " a", "start": 343.82, "end": 344.0, "probability": 0.9949306845664978}, {"word": " certain", "start": 344.0, "end": 344.18, "probability": 0.9980814456939697}, {"word": " degree", "start": 344.18, "end": 344.46, "probability": 0.999301552772522}, {"word": " of", "start": 344.46, "end": 344.7, "probability": 0.9993768334388733}, {"word": " trust", "start": 344.7, "end": 345.04, "probability": 0.9964234232902527}]}, {"id": 67, "seek": 32552, "start": 346.76000000000005, "end": 349.38, "text": " to the knowledge it obtains from its neighbors.", "tokens": [51383, 284, 262, 3725, 340, 909, 12143, 422, 663, 12020, 13, 51559], "temperature": 0.0, "avg_logprob": -0.09193602204322815, "compression_ratio": 1.5720524017467248, "no_speech_prob": 0.0036252900026738644, "words": [{"word": " to", "start": 346.76000000000005, "end": 347.28000000000003, "probability": 0.665306568145752}, {"word": " the", "start": 347.28000000000003, "end": 347.8, "probability": 0.9883032441139221}, {"word": " knowledge", "start": 347.8, "end": 348.02, "probability": 0.9988881945610046}, {"word": " it", "start": 348.02, "end": 348.22, "probability": 0.9605777263641357}, {"word": " obtains", "start": 348.22, "end": 348.66, "probability": 0.9762499630451202}, {"word": " from", "start": 348.66, "end": 348.96, "probability": 0.9960441589355469}, {"word": " its", "start": 348.96, "end": 349.12, "probability": 0.990523636341095}, {"word": " neighbors.", "start": 349.12, "end": 349.38, "probability": 0.935745358467102}]}, {"id": 68, "seek": 34938, "start": 350.32, "end": 357.0, "text": " I think I train a neural network locally on my compute using my data collected by my sensors.", "tokens": [50363, 314, 892, 314, 4512, 257, 17019, 3127, 15726, 319, 616, 24061, 1262, 616, 1366, 7723, 416, 616, 15736, 13, 50719], "temperature": 0.0, "avg_logprob": -0.21545241095803, "compression_ratio": 1.5672268907563025, "no_speech_prob": 0.011322530917823315, "words": [{"word": " I", "start": 350.32, "end": 350.92, "probability": 0.310637503862381}, {"word": " think", "start": 350.92, "end": 351.12, "probability": 0.8631237149238586}, {"word": " I", "start": 351.12, "end": 352.04, "probability": 0.7853065729141235}, {"word": " train", "start": 352.04, "end": 352.62, "probability": 0.7116039991378784}, {"word": " a", "start": 352.62, "end": 352.76, "probability": 0.646289050579071}, {"word": " neural", "start": 352.76, "end": 352.94, "probability": 0.9782800078392029}, {"word": " network", "start": 352.94, "end": 353.26, "probability": 0.9987596273422241}, {"word": " locally", "start": 353.26, "end": 353.66, "probability": 0.9914504885673523}, {"word": " on", "start": 353.66, "end": 353.98, "probability": 0.9519165754318237}, {"word": " my", "start": 353.98, "end": 354.18, "probability": 0.9985432624816895}, {"word": " compute", "start": 354.18, "end": 354.52, "probability": 0.9165999293327332}, {"word": " using", "start": 354.52, "end": 354.96, "probability": 0.932880699634552}, {"word": " my", "start": 354.96, "end": 355.24, "probability": 0.9964549541473389}, {"word": " data", "start": 355.24, "end": 355.56, "probability": 0.9994632601737976}, {"word": " collected", "start": 355.56, "end": 356.14, "probability": 0.7829036116600037}, {"word": " by", "start": 356.14, "end": 356.42, "probability": 0.9973176121711731}, {"word": " my", "start": 356.42, "end": 356.62, "probability": 0.9683747291564941}, {"word": " sensors.", "start": 356.62, "end": 357.0, "probability": 0.990628182888031}]}, {"id": 69, "seek": 34938, "start": 357.6, "end": 363.84, "text": " This could be my phone's GPS, my fitness wearable, accelerometer, camera, security system, etc.", "tokens": [50719, 770, 714, 307, 616, 3072, 338, 15472, 11, 616, 13547, 39092, 11, 8320, 15635, 11, 4676, 11, 2324, 1080, 11, 3503, 13, 51055], "temperature": 0.0, "avg_logprob": -0.21545241095803, "compression_ratio": 1.5672268907563025, "no_speech_prob": 0.011322530917823315, "words": [{"word": " This", "start": 357.6, "end": 357.76, "probability": 0.994207501411438}, {"word": " could", "start": 357.76, "end": 357.94, "probability": 0.991707980632782}, {"word": " be", "start": 357.94, "end": 358.26, "probability": 0.9995133876800537}, {"word": " my", "start": 358.26, "end": 358.52, "probability": 0.9932576417922974}, {"word": " phone's", "start": 358.52, "end": 358.86, "probability": 0.96749147772789}, {"word": " GPS,", "start": 358.86, "end": 359.28, "probability": 0.9971674084663391}, {"word": " my", "start": 359.84, "end": 359.96, "probability": 0.9940608143806458}, {"word": " fitness", "start": 359.96, "end": 360.32, "probability": 0.9980923533439636}, {"word": " wearable,", "start": 360.32, "end": 360.76, "probability": 0.9705091118812561}, {"word": " accelerometer,", "start": 361.06, "end": 361.68, "probability": 0.9927383661270142}, {"word": " camera,", "start": 361.96, "end": 362.26, "probability": 0.9981587529182434}, {"word": " security", "start": 362.44, "end": 362.76, "probability": 0.9989012479782104}, {"word": " system,", "start": 362.76, "end": 363.32, "probability": 0.98914635181427}, {"word": " etc.", "start": 363.6, "end": 363.84, "probability": 0.4133351445198059}]}, {"id": 70, "seek": 34938, "start": 364.58, "end": 367.44, "text": " And then I share that knowledge with others in the network.", "tokens": [51083, 843, 788, 314, 2648, 326, 3725, 351, 1854, 287, 262, 3127, 13, 51227], "temperature": 0.0, "avg_logprob": -0.21545241095803, "compression_ratio": 1.5672268907563025, "no_speech_prob": 0.011322530917823315, "words": [{"word": " And", "start": 364.58, "end": 364.82, "probability": 0.839059591293335}, {"word": " then", "start": 364.82, "end": 364.94, "probability": 0.995032548904419}, {"word": " I", "start": 364.94, "end": 365.12, "probability": 0.972540020942688}, {"word": " share", "start": 365.12, "end": 365.32, "probability": 0.9962445497512817}, {"word": " that", "start": 365.32, "end": 365.54, "probability": 0.999160885810852}, {"word": " knowledge", "start": 365.54, "end": 365.9, "probability": 0.9995438456535339}, {"word": " with", "start": 365.9, "end": 366.38, "probability": 0.998775064945221}, {"word": " others", "start": 366.38, "end": 366.7, "probability": 0.9976391792297363}, {"word": " in", "start": 366.7, "end": 367.04, "probability": 0.9980794191360474}, {"word": " the", "start": 367.04, "end": 367.14, "probability": 0.999175488948822}, {"word": " network.", "start": 367.14, "end": 367.44, "probability": 0.9983306527137756}]}, {"id": 71, "seek": 34938, "start": 368.88, "end": 372.18, "text": " Unfortunately, the same math that allows you to turn data into parameters", "tokens": [51291, 8989, 11, 262, 976, 10688, 326, 3578, 345, 284, 1210, 1366, 656, 10007, 51503], "temperature": 0.0, "avg_logprob": -0.21545241095803, "compression_ratio": 1.5672268907563025, "no_speech_prob": 0.011322530917823315, "words": [{"word": " Unfortunately,", "start": 368.88, "end": 369.4, "probability": 0.993245542049408}, {"word": " the", "start": 369.86, "end": 370.02, "probability": 0.9988877177238464}, {"word": " same", "start": 370.02, "end": 370.28, "probability": 0.9987755417823792}, {"word": " math", "start": 370.28, "end": 370.58, "probability": 0.9778743386268616}, {"word": " that", "start": 370.58, "end": 370.76, "probability": 0.9941611886024475}, {"word": " allows", "start": 370.76, "end": 371.04, "probability": 0.9987021684646606}, {"word": " you", "start": 371.04, "end": 371.28, "probability": 0.9991521835327148}, {"word": " to", "start": 371.28, "end": 371.4, "probability": 0.999194324016571}, {"word": " turn", "start": 371.4, "end": 371.6, "probability": 0.9974480867385864}, {"word": " data", "start": 371.6, "end": 371.9, "probability": 0.9985802173614502}, {"word": " into", "start": 371.9, "end": 372.16, "probability": 0.995537519454956}, {"word": " parameters", "start": 372.16, "end": 372.46000000000004, "probability": 0.9971204996109009}]}, {"id": 72, "seek": 34938, "start": 372.76, "end": 375.96, "text": " can turn parameters into data, your private data.", "tokens": [51503, 460, 1210, 10007, 656, 1366, 11, 534, 2839, 1366, 13, 51655], "temperature": 0.0, "avg_logprob": -0.21545241095803, "compression_ratio": 1.5672268907563025, "no_speech_prob": 0.011322530917823315, "words": [{"word": " can", "start": 372.76, "end": 373.24, "probability": 0.8584262132644653}, {"word": " turn", "start": 373.24, "end": 373.46, "probability": 0.9982274174690247}, {"word": " parameters", "start": 373.46, "end": 374.0, "probability": 0.9973913431167603}, {"word": " into", "start": 374.0, "end": 374.56, "probability": 0.9916268587112427}, {"word": " data,", "start": 374.56, "end": 374.88, "probability": 0.9076462984085083}, {"word": " your", "start": 375.16, "end": 375.32, "probability": 0.991759717464447}, {"word": " private", "start": 375.32, "end": 375.66, "probability": 0.9985654950141907}, {"word": " data.", "start": 375.66, "end": 375.96, "probability": 0.9989064931869507}]}, {"id": 73, "seek": 37596, "start": 376.59999999999997, "end": 381.24, "text": " This can be mitigated in several ways, including just sharing parameters less frequently.", "tokens": [50395, 770, 460, 307, 10255, 26963, 287, 1811, 2842, 11, 1390, 655, 7373, 10007, 1342, 6777, 13, 50647], "temperature": 0.0, "avg_logprob": -0.1307221674451641, "compression_ratio": 1.6926406926406927, "no_speech_prob": 0.013112956658005714, "words": [{"word": " This", "start": 376.58000000000004, "end": 376.86, "probability": 0.8441006541252136}, {"word": " can", "start": 376.86, "end": 377.04, "probability": 0.9853581190109253}, {"word": " be", "start": 377.04, "end": 377.18, "probability": 0.9985054731369019}, {"word": " mitigated", "start": 377.18, "end": 377.64, "probability": 0.9954836368560791}, {"word": " in", "start": 377.64, "end": 377.94, "probability": 0.9958840012550354}, {"word": " several", "start": 377.94, "end": 378.18, "probability": 0.9984042048454285}, {"word": " ways,", "start": 378.18, "end": 378.68, "probability": 0.9988382458686829}, {"word": " including", "start": 379.22, "end": 379.34, "probability": 0.9963544607162476}, {"word": " just", "start": 379.34, "end": 379.76, "probability": 0.9623169302940369}, {"word": " sharing", "start": 379.76, "end": 380.12, "probability": 0.9979891777038574}, {"word": " parameters", "start": 380.12, "end": 380.56, "probability": 0.989963710308075}, {"word": " less", "start": 380.56, "end": 380.8, "probability": 0.9919446110725403}, {"word": " frequently.", "start": 380.8, "end": 381.24, "probability": 0.9969984292984009}]}, {"id": 74, "seek": 37596, "start": 381.78, "end": 386.0, "text": " This is an ongoing area of research, but in general, this problem can be dealt with.", "tokens": [50647, 770, 318, 281, 7044, 1989, 286, 2267, 11, 475, 287, 2276, 11, 428, 1917, 460, 307, 11829, 351, 13, 50851], "temperature": 0.0, "avg_logprob": -0.1307221674451641, "compression_ratio": 1.6926406926406927, "no_speech_prob": 0.013112956658005714, "words": [{"word": " This", "start": 381.78, "end": 381.94, "probability": 0.9962581396102905}, {"word": " is", "start": 381.94, "end": 382.06, "probability": 0.9956064820289612}, {"word": " an", "start": 382.06, "end": 382.2, "probability": 0.9969586133956909}, {"word": " ongoing", "start": 382.2, "end": 382.74, "probability": 0.9869896769523621}, {"word": " area", "start": 382.74, "end": 383.1, "probability": 0.997092604637146}, {"word": " of", "start": 383.1, "end": 383.26, "probability": 0.9986605644226074}, {"word": " research,", "start": 383.26, "end": 383.56, "probability": 0.9987850785255432}, {"word": " but", "start": 383.68, "end": 383.84, "probability": 0.9981496334075928}, {"word": " in", "start": 383.84, "end": 384.0, "probability": 0.9692034721374512}, {"word": " general,", "start": 384.0, "end": 384.36, "probability": 0.9996423721313477}, {"word": " this", "start": 384.42, "end": 384.7, "probability": 0.9989370703697205}, {"word": " problem", "start": 384.7, "end": 385.0, "probability": 0.9990792274475098}, {"word": " can", "start": 385.0, "end": 385.24, "probability": 0.9992352724075317}, {"word": " be", "start": 385.24, "end": 385.36, "probability": 0.9997554421424866}, {"word": " dealt", "start": 385.36, "end": 385.52, "probability": 0.9994304776191711}, {"word": " with.", "start": 385.52, "end": 386.0, "probability": 0.9989141225814819}]}, {"id": 75, "seek": 37596, "start": 386.96, "end": 388.52, "text": " It's a losing game.", "tokens": [50899, 632, 338, 257, 6078, 983, 13, 50987], "temperature": 0.0, "avg_logprob": -0.1307221674451641, "compression_ratio": 1.6926406926406927, "no_speech_prob": 0.013112956658005714, "words": [{"word": " It's", "start": 386.96, "end": 387.12, "probability": 0.9481134414672852}, {"word": " a", "start": 387.12, "end": 387.88, "probability": 0.8930308818817139}, {"word": " losing", "start": 387.88, "end": 388.08, "probability": 0.9687107801437378}, {"word": " game.", "start": 388.08, "end": 388.52, "probability": 0.9985432624816895}]}, {"id": 76, "seek": 37596, "start": 390.74, "end": 394.64, "text": " Yes, there's a lot of research that's been done into this.", "tokens": [51119, 3363, 11, 612, 338, 257, 1256, 286, 2267, 326, 338, 587, 1760, 656, 428, 13, 51303], "temperature": 0.0, "avg_logprob": -0.1307221674451641, "compression_ratio": 1.6926406926406927, "no_speech_prob": 0.013112956658005714, "words": [{"word": " Yes,", "start": 390.74, "end": 391.3, "probability": 0.8958091735839844}, {"word": " there's", "start": 392.18, "end": 392.46, "probability": 0.9951250851154327}, {"word": " a", "start": 392.46, "end": 392.62, "probability": 0.9935441017150879}, {"word": " lot", "start": 392.62, "end": 393.06, "probability": 0.9609144330024719}, {"word": " of", "start": 393.06, "end": 393.16, "probability": 0.9990054965019226}, {"word": " research", "start": 393.16, "end": 393.4, "probability": 0.9994246959686279}, {"word": " that's", "start": 393.4, "end": 393.68, "probability": 0.9960231184959412}, {"word": " been", "start": 393.68, "end": 393.84, "probability": 0.9991266131401062}, {"word": " done", "start": 393.84, "end": 394.06, "probability": 0.998833954334259}, {"word": " into", "start": 394.06, "end": 394.36, "probability": 0.9968780279159546}, {"word": " this.", "start": 394.36, "end": 394.64, "probability": 0.9996626377105713}]}, {"id": 77, "seek": 37596, "start": 394.96, "end": 397.54, "text": " And in general, it is a losing game for the attackers.", "tokens": [51303, 843, 287, 2276, 11, 340, 318, 257, 6078, 983, 329, 262, 16391, 13, 51447], "temperature": 0.0, "avg_logprob": -0.1307221674451641, "compression_ratio": 1.6926406926406927, "no_speech_prob": 0.013112956658005714, "words": [{"word": " And", "start": 394.96, "end": 395.02, "probability": 0.9318856596946716}, {"word": " in", "start": 395.02, "end": 395.18, "probability": 0.9749068021774292}, {"word": " general,", "start": 395.18, "end": 395.56, "probability": 0.9996271133422852}, {"word": " it", "start": 395.62, "end": 395.96, "probability": 0.9992192983627319}, {"word": " is", "start": 395.96, "end": 396.12, "probability": 0.9985217452049255}, {"word": " a", "start": 396.12, "end": 396.24, "probability": 0.9979804158210754}, {"word": " losing", "start": 396.24, "end": 396.42, "probability": 0.9992314577102661}, {"word": " game", "start": 396.42, "end": 396.72, "probability": 0.9992720484733582}, {"word": " for", "start": 396.72, "end": 396.94, "probability": 0.9989542961120605}, {"word": " the", "start": 396.94, "end": 397.12, "probability": 0.9993991851806641}, {"word": " attackers.", "start": 397.12, "end": 397.54, "probability": 0.9341522455215454}]}, {"id": 78, "seek": 37596, "start": 398.46, "end": 403.0, "text": " The ways that you can extract it that have been done, it's very contrived, very...", "tokens": [51495, 383, 2842, 326, 345, 460, 7925, 340, 326, 423, 587, 1760, 11, 340, 338, 845, 542, 36207, 11, 845, 986, 51715], "temperature": 0.0, "avg_logprob": -0.1307221674451641, "compression_ratio": 1.6926406926406927, "no_speech_prob": 0.013112956658005714, "words": [{"word": " The", "start": 398.46, "end": 398.46, "probability": 0.9327248930931091}, {"word": " ways", "start": 398.46, "end": 399.08, "probability": 0.980635941028595}, {"word": " that", "start": 399.08, "end": 399.28, "probability": 0.9213863611221313}, {"word": " you", "start": 399.28, "end": 399.34, "probability": 0.9994421601295471}, {"word": " can", "start": 399.34, "end": 399.48, "probability": 0.9992551207542419}, {"word": " extract", "start": 399.48, "end": 399.86, "probability": 0.9987128973007202}, {"word": " it", "start": 399.86, "end": 400.04, "probability": 0.993419349193573}, {"word": " that", "start": 400.04, "end": 400.16, "probability": 0.9610403776168823}, {"word": " have", "start": 400.16, "end": 400.28, "probability": 0.9895181059837341}, {"word": " been", "start": 400.28, "end": 400.46, "probability": 0.998753547668457}, {"word": " done,", "start": 400.46, "end": 400.76, "probability": 0.9988709092140198}, {"word": " it's", "start": 400.94, "end": 401.24, "probability": 0.9959290325641632}, {"word": " very", "start": 401.24, "end": 401.54, "probability": 0.9822502732276917}, {"word": " contrived,", "start": 401.54, "end": 402.48, "probability": 0.9976999759674072}, {"word": " very...", "start": 402.84, "end": 403.11999999999995, "probability": 0.6932741850614548}]}, {"id": 79, "seek": 40596, "start": 405.96, "end": 407.5, "text": " not realistic scenarios.", "tokens": [50363, 407, 12653, 13858, 13, 50439], "temperature": 0.0, "avg_logprob": -0.1072525691986084, "compression_ratio": 1.60546875, "no_speech_prob": 0.24872173368930817, "words": [{"word": " not", "start": 405.96, "end": 406.28, "probability": 0.3639047145843506}, {"word": " realistic", "start": 406.28, "end": 406.84, "probability": 0.9479350447654724}, {"word": " scenarios.", "start": 406.84, "end": 407.5, "probability": 0.896327018737793}]}, {"id": 80, "seek": 40596, "start": 408.7, "end": 413.3, "text": " But talk to me after this, if you want to discuss this more, that the next one is a bit more of a", "tokens": [50495, 887, 1561, 284, 502, 706, 428, 11, 611, 345, 765, 284, 2112, 428, 517, 11, 326, 262, 1306, 530, 318, 257, 1643, 517, 286, 257, 50727], "temperature": 0.0, "avg_logprob": -0.1072525691986084, "compression_ratio": 1.60546875, "no_speech_prob": 0.24872173368930817, "words": [{"word": " But", "start": 408.7, "end": 408.86, "probability": 0.6501331329345703}, {"word": " talk", "start": 408.86, "end": 409.4, "probability": 0.8600082993507385}, {"word": " to", "start": 409.4, "end": 409.56, "probability": 0.9972261786460876}, {"word": " me", "start": 409.56, "end": 409.66, "probability": 0.9986120462417603}, {"word": " after", "start": 409.66, "end": 409.86, "probability": 0.988895058631897}, {"word": " this,", "start": 409.86, "end": 410.04, "probability": 0.9913952946662903}, {"word": " if", "start": 410.04, "end": 410.16, "probability": 0.9804915189743042}, {"word": " you", "start": 410.16, "end": 410.28, "probability": 0.9979670643806458}, {"word": " want", "start": 410.28, "end": 410.38, "probability": 0.867475152015686}, {"word": " to", "start": 410.38, "end": 410.48, "probability": 0.9975651502609253}, {"word": " discuss", "start": 410.48, "end": 410.72, "probability": 0.9975777268409729}, {"word": " this", "start": 410.72, "end": 410.96, "probability": 0.9797065258026123}, {"word": " more,", "start": 410.96, "end": 411.2, "probability": 0.9769779443740845}, {"word": " that", "start": 411.28, "end": 411.46, "probability": 0.7386642098426819}, {"word": " the", "start": 411.46, "end": 411.84, "probability": 0.982738196849823}, {"word": " next", "start": 411.84, "end": 412.0, "probability": 0.9983736276626587}, {"word": " one", "start": 412.0, "end": 412.2, "probability": 0.9964403510093689}, {"word": " is", "start": 412.2, "end": 412.36, "probability": 0.9945234060287476}, {"word": " a", "start": 412.36, "end": 412.7, "probability": 0.9002167582511902}, {"word": " bit", "start": 412.7, "end": 412.84, "probability": 0.9900674819946289}, {"word": " more", "start": 412.84, "end": 413.08, "probability": 0.9989234805107117}, {"word": " of", "start": 413.08, "end": 413.18, "probability": 0.997260570526123}, {"word": " a", "start": 413.18, "end": 413.3, "probability": 0.9988279938697815}]}, {"id": 81, "seek": 40596, "start": 413.3, "end": 419.32, "text": " problem. The other major consideration is malicious actors. For instance, I could take a", "tokens": [50727, 1917, 13, 383, 584, 1688, 9110, 318, 17412, 10544, 13, 1114, 4554, 11, 314, 714, 1011, 257, 51023], "temperature": 0.0, "avg_logprob": -0.1072525691986084, "compression_ratio": 1.60546875, "no_speech_prob": 0.24872173368930817, "words": [{"word": " problem.", "start": 413.3, "end": 413.6, "probability": 0.9987422823905945}, {"word": " The", "start": 414.44, "end": 414.58, "probability": 0.9758697748184204}, {"word": " other", "start": 414.58, "end": 414.88, "probability": 0.9993763566017151}, {"word": " major", "start": 414.88, "end": 415.4, "probability": 0.999101996421814}, {"word": " consideration", "start": 415.4, "end": 415.94, "probability": 0.9936955571174622}, {"word": " is", "start": 415.94, "end": 416.32, "probability": 0.9976564645767212}, {"word": " malicious", "start": 416.32, "end": 416.8, "probability": 0.9648570418357849}, {"word": " actors.", "start": 416.8, "end": 417.36, "probability": 0.998206615447998}, {"word": " For", "start": 417.92, "end": 418.02, "probability": 0.9963114857673645}, {"word": " instance,", "start": 418.02, "end": 418.26, "probability": 0.9996366500854492}, {"word": " I", "start": 418.72, "end": 418.86, "probability": 0.9940919280052185}, {"word": " could", "start": 418.86, "end": 419.02, "probability": 0.9832344055175781}, {"word": " take", "start": 419.02, "end": 419.18, "probability": 0.9995396137237549}, {"word": " a", "start": 419.18, "end": 419.32, "probability": 0.9996782541275024}]}, {"id": 82, "seek": 40596, "start": 419.32, "end": 425.46, "text": " bunch of photos of myself, label them as not a burglar, contribute to a federated learning network,", "tokens": [51023, 7684, 286, 5205, 286, 3589, 11, 6167, 606, 355, 407, 257, 15601, 21681, 11, 8676, 284, 257, 28062, 515, 4673, 3127, 11, 51343], "temperature": 0.0, "avg_logprob": -0.1072525691986084, "compression_ratio": 1.60546875, "no_speech_prob": 0.24872173368930817, "words": [{"word": " bunch", "start": 419.32, "end": 419.44, "probability": 0.9996813535690308}, {"word": " of", "start": 419.44, "end": 419.6, "probability": 0.9995668530464172}, {"word": " photos", "start": 419.6, "end": 419.86, "probability": 0.9990726709365845}, {"word": " of", "start": 419.86, "end": 420.04, "probability": 0.9993902444839478}, {"word": " myself,", "start": 420.04, "end": 420.42, "probability": 0.8740622401237488}, {"word": " label", "start": 420.94, "end": 421.22, "probability": 0.9926512837409973}, {"word": " them", "start": 421.22, "end": 421.54, "probability": 0.9990547299385071}, {"word": " as", "start": 421.54, "end": 421.74, "probability": 0.9955244064331055}, {"word": " not", "start": 421.74, "end": 422.02, "probability": 0.9059353470802307}, {"word": " a", "start": 422.02, "end": 422.26, "probability": 0.982860803604126}, {"word": " burglar,", "start": 422.26, "end": 422.6, "probability": 0.9931640923023224}, {"word": " contribute", "start": 423.58, "end": 423.9, "probability": 0.9914152026176453}, {"word": " to", "start": 423.9, "end": 424.24, "probability": 0.9995661377906799}, {"word": " a", "start": 424.24, "end": 424.34, "probability": 0.9980996251106262}, {"word": " federated", "start": 424.34, "end": 424.68, "probability": 0.9694684743881226}, {"word": " learning", "start": 424.68, "end": 424.98, "probability": 0.9872888326644897}, {"word": " network,", "start": 424.98, "end": 425.46, "probability": 0.9984583854675293}]}, {"id": 83, "seek": 40596, "start": 425.7, "end": 431.58, "text": " and suddenly people's home security system labels me as not a burglar because I lied to the network", "tokens": [51343, 290, 6451, 661, 338, 1363, 2324, 1080, 14722, 502, 355, 407, 257, 15601, 21681, 780, 314, 19837, 284, 262, 3127, 51643], "temperature": 0.0, "avg_logprob": -0.1072525691986084, "compression_ratio": 1.60546875, "no_speech_prob": 0.24872173368930817, "words": [{"word": " and", "start": 425.7, "end": 425.86, "probability": 0.9981830716133118}, {"word": " suddenly", "start": 425.86, "end": 426.18, "probability": 0.9916960597038269}, {"word": " people's", "start": 426.18, "end": 426.72, "probability": 0.8721074759960175}, {"word": " home", "start": 426.72, "end": 426.86, "probability": 0.9787217974662781}, {"word": " security", "start": 426.86, "end": 427.26, "probability": 0.983899712562561}, {"word": " system", "start": 427.26, "end": 427.74, "probability": 0.9917296767234802}, {"word": " labels", "start": 427.74, "end": 428.18, "probability": 0.9906903505325317}, {"word": " me", "start": 428.18, "end": 428.54, "probability": 0.9989694356918335}, {"word": " as", "start": 428.54, "end": 429.08, "probability": 0.9976873397827148}, {"word": " not", "start": 429.08, "end": 429.6, "probability": 0.9113854765892029}, {"word": " a", "start": 429.6, "end": 429.8, "probability": 0.9972847700119019}, {"word": " burglar", "start": 429.8, "end": 430.12, "probability": 0.9975137114524841}, {"word": " because", "start": 430.12, "end": 430.58, "probability": 0.52040696144104}, {"word": " I", "start": 430.58, "end": 430.76, "probability": 0.9983903169631958}, {"word": " lied", "start": 430.76, "end": 431.02, "probability": 0.9954684972763062}, {"word": " to", "start": 431.02, "end": 431.22, "probability": 0.9993544220924377}, {"word": " the", "start": 431.22, "end": 431.32, "probability": 0.9985687732696533}, {"word": " network", "start": 431.32, "end": 431.58, "probability": 0.9988563060760498}]}, {"id": 84, "seek": 43158, "start": 431.58, "end": 436.4, "text": " and told it that things that look like me are not burglars. Basically, we need to develop", "tokens": [50363, 290, 1297, 340, 326, 1243, 326, 804, 588, 502, 389, 407, 26574, 945, 13, 20759, 11, 356, 761, 284, 1205, 50607], "temperature": 0.0, "avg_logprob": -0.07553275993892125, "compression_ratio": 1.5909090909090908, "no_speech_prob": 0.10035055875778198, "words": [{"word": " and", "start": 431.58, "end": 431.76, "probability": 0.7632679343223572}, {"word": " told", "start": 431.76, "end": 431.92, "probability": 0.9899015426635742}, {"word": " it", "start": 431.92, "end": 432.06, "probability": 0.9009045362472534}, {"word": " that", "start": 432.06, "end": 432.22, "probability": 0.9791023135185242}, {"word": " things", "start": 432.22, "end": 432.44, "probability": 0.881549596786499}, {"word": " that", "start": 432.44, "end": 432.62, "probability": 0.9904159307479858}, {"word": " look", "start": 432.62, "end": 432.8, "probability": 0.9462838768959045}, {"word": " like", "start": 432.8, "end": 432.98, "probability": 0.9984207153320312}, {"word": " me", "start": 432.98, "end": 433.2, "probability": 0.6738023161888123}, {"word": " are", "start": 433.2, "end": 433.46, "probability": 0.9390302300453186}, {"word": " not", "start": 433.46, "end": 433.66, "probability": 0.9978729486465454}, {"word": " burglars.", "start": 433.66, "end": 434.16, "probability": 0.9924319386482239}, {"word": " Basically,", "start": 435.24, "end": 435.24, "probability": 0.9682856202125549}, {"word": " we", "start": 435.24, "end": 435.44, "probability": 0.997708797454834}, {"word": " need", "start": 435.44, "end": 435.9, "probability": 0.911810576915741}, {"word": " to", "start": 435.9, "end": 436.06, "probability": 0.9974169731140137}, {"word": " develop", "start": 436.06, "end": 436.4, "probability": 0.997962236404419}]}, {"id": 85, "seek": 43158, "start": 436.4, "end": 443.22, "text": " mechanisms for detecting and ignoring liars. This is also a big area of research.", "tokens": [50607, 11701, 329, 31521, 290, 15482, 7649, 945, 13, 770, 318, 635, 257, 1263, 1989, 286, 2267, 13, 50947], "temperature": 0.0, "avg_logprob": -0.07553275993892125, "compression_ratio": 1.5909090909090908, "no_speech_prob": 0.10035055875778198, "words": [{"word": " mechanisms", "start": 436.4, "end": 436.98, "probability": 0.9940967559814453}, {"word": " for", "start": 436.98, "end": 437.4, "probability": 0.9971404075622559}, {"word": " detecting", "start": 437.4, "end": 437.82, "probability": 0.9979876279830933}, {"word": " and", "start": 437.82, "end": 438.06, "probability": 0.9943935871124268}, {"word": " ignoring", "start": 438.06, "end": 438.4, "probability": 0.999254047870636}, {"word": " liars.", "start": 438.4, "end": 438.9, "probability": 0.9803830981254578}, {"word": " This", "start": 439.4, "end": 439.52, "probability": 0.9941258430480957}, {"word": " is", "start": 439.52, "end": 439.66, "probability": 0.9984925985336304}, {"word": " also", "start": 439.66, "end": 440.08, "probability": 0.9941495656967163}, {"word": " a", "start": 440.08, "end": 440.96, "probability": 0.9909006357192993}, {"word": " big", "start": 440.96, "end": 441.9, "probability": 0.9384392499923706}, {"word": " area", "start": 441.9, "end": 442.32, "probability": 0.9976180195808411}, {"word": " of", "start": 442.32, "end": 442.78, "probability": 0.9988903403282166}, {"word": " research.", "start": 442.78, "end": 443.22, "probability": 0.9995059967041016}]}, {"id": 86, "seek": 43158, "start": 444.44, "end": 451.16, "text": " I think it's certainly this is the major hard point of decentralized federated learning,", "tokens": [50995, 314, 892, 340, 338, 3729, 428, 318, 262, 1688, 1327, 966, 286, 26512, 28062, 515, 4673, 11, 51351], "temperature": 0.0, "avg_logprob": -0.07553275993892125, "compression_ratio": 1.5909090909090908, "no_speech_prob": 0.10035055875778198, "words": [{"word": " I", "start": 444.44, "end": 444.44, "probability": 0.8345597982406616}, {"word": " think", "start": 444.44, "end": 444.62, "probability": 0.9977414608001709}, {"word": " it's", "start": 444.62, "end": 444.86, "probability": 0.8467331826686859}, {"word": " certainly", "start": 444.86, "end": 445.5, "probability": 0.9202117323875427}, {"word": " this", "start": 445.5, "end": 445.88, "probability": 0.4289034307003021}, {"word": " is", "start": 445.88, "end": 446.06, "probability": 0.9932467937469482}, {"word": " the", "start": 446.06, "end": 446.32, "probability": 0.9975036978721619}, {"word": " major", "start": 446.32, "end": 446.92, "probability": 0.9962292313575745}, {"word": " hard", "start": 446.92, "end": 448.52, "probability": 0.8853330612182617}, {"word": " point", "start": 448.52, "end": 448.9, "probability": 0.8771793246269226}, {"word": " of", "start": 448.9, "end": 449.56, "probability": 0.9946281313896179}, {"word": " decentralized", "start": 449.56, "end": 450.3, "probability": 0.973505437374115}, {"word": " federated", "start": 450.3, "end": 450.8, "probability": 0.9398776590824127}, {"word": " learning,", "start": 450.8, "end": 451.16, "probability": 0.9988985061645508}]}, {"id": 87, "seek": 43158, "start": 451.36, "end": 456.4, "text": " but I do think that it's not an obstacle that cannot be overcome. I think we can overcome", "tokens": [51351, 475, 314, 466, 892, 326, 340, 338, 407, 281, 22007, 326, 2314, 307, 10980, 13, 314, 892, 356, 460, 10980, 51603], "temperature": 0.0, "avg_logprob": -0.07553275993892125, "compression_ratio": 1.5909090909090908, "no_speech_prob": 0.10035055875778198, "words": [{"word": " but", "start": 451.36, "end": 451.68, "probability": 0.9968540072441101}, {"word": " I", "start": 451.68, "end": 452.16, "probability": 0.9863138198852539}, {"word": " do", "start": 452.16, "end": 452.38, "probability": 0.9986224174499512}, {"word": " think", "start": 452.38, "end": 452.58, "probability": 0.9985806941986084}, {"word": " that", "start": 452.58, "end": 452.82, "probability": 0.9888888001441956}, {"word": " it's", "start": 452.82, "end": 453.3, "probability": 0.9959942102432251}, {"word": " not", "start": 453.3, "end": 453.64, "probability": 0.9986911416053772}, {"word": " an", "start": 453.64, "end": 453.82, "probability": 0.9985687732696533}, {"word": " obstacle", "start": 453.82, "end": 454.26, "probability": 0.9995686411857605}, {"word": " that", "start": 454.26, "end": 454.58, "probability": 0.994708776473999}, {"word": " cannot", "start": 454.58, "end": 454.9, "probability": 0.9786596894264221}, {"word": " be", "start": 454.9, "end": 455.12, "probability": 0.9971399307250977}, {"word": " overcome.", "start": 455.12, "end": 455.46, "probability": 0.999040424823761}, {"word": " I", "start": 455.62, "end": 455.64, "probability": 0.9961049556732178}, {"word": " think", "start": 455.64, "end": 455.74, "probability": 0.9996111989021301}, {"word": " we", "start": 455.74, "end": 455.86, "probability": 0.9983977675437927}, {"word": " can", "start": 455.86, "end": 456.06, "probability": 0.9659287333488464}, {"word": " overcome", "start": 456.06, "end": 456.4, "probability": 0.9314948916435242}]}, {"id": 88, "seek": 45640, "start": 456.4, "end": 463.32, "text": " this obstacle. So in conclusion, internet networks cannot handle the data throughput to", "tokens": [50363, 428, 22007, 13, 1406, 287, 7664, 11, 5230, 7686, 2314, 5412, 262, 1366, 41997, 284, 50719], "temperature": 0.0, "avg_logprob": -0.07136473140201054, "compression_ratio": 1.6009174311926606, "no_speech_prob": 0.002968456130474806, "words": [{"word": " this", "start": 456.4, "end": 456.58, "probability": 0.22625237703323364}, {"word": " obstacle.", "start": 456.58, "end": 457.04, "probability": 0.9830818772315979}, {"word": " So", "start": 458.3, "end": 458.3, "probability": 0.8269454836845398}, {"word": " in", "start": 458.3, "end": 458.44, "probability": 0.7621249556541443}, {"word": " conclusion,", "start": 458.44, "end": 458.92, "probability": 0.9978475570678711}, {"word": " internet", "start": 460.08, "end": 460.16, "probability": 0.6290079355239868}, {"word": " networks", "start": 460.16, "end": 460.62, "probability": 0.9964723587036133}, {"word": " cannot", "start": 460.62, "end": 460.88, "probability": 0.98658686876297}, {"word": " handle", "start": 460.88, "end": 461.32, "probability": 0.9976968169212341}, {"word": " the", "start": 461.32, "end": 461.94, "probability": 0.9885528087615967}, {"word": " data", "start": 461.94, "end": 462.56, "probability": 0.9983001351356506}, {"word": " throughput", "start": 462.56, "end": 463.04, "probability": 0.9430832862854004}, {"word": " to", "start": 463.04, "end": 463.32, "probability": 0.9813855886459351}]}, {"id": 89, "seek": 45640, "start": 463.32, "end": 468.52, "text": " centralize all data. That means more data is available for federated learning frameworks,", "tokens": [50719, 4318, 1096, 477, 1366, 13, 1320, 1724, 517, 1366, 318, 1695, 329, 28062, 515, 4673, 29251, 11, 50975], "temperature": 0.0, "avg_logprob": -0.07136473140201054, "compression_ratio": 1.6009174311926606, "no_speech_prob": 0.002968456130474806, "words": [{"word": " centralize", "start": 463.32, "end": 464.18, "probability": 0.9778708815574646}, {"word": " all", "start": 464.18, "end": 464.64, "probability": 0.9810457229614258}, {"word": " data.", "start": 464.64, "end": 464.98, "probability": 0.9940511584281921}, {"word": " That", "start": 465.38, "end": 465.38, "probability": 0.9829401969909668}, {"word": " means", "start": 465.38, "end": 465.56, "probability": 0.9988308548927307}, {"word": " more", "start": 465.56, "end": 465.9, "probability": 0.9861118793487549}, {"word": " data", "start": 465.9, "end": 466.16, "probability": 0.9982215762138367}, {"word": " is", "start": 466.16, "end": 466.42, "probability": 0.9933338761329651}, {"word": " available", "start": 466.42, "end": 466.82, "probability": 0.9990139007568359}, {"word": " for", "start": 466.82, "end": 467.4, "probability": 0.9972355961799622}, {"word": " federated", "start": 467.4, "end": 467.84, "probability": 0.903256744146347}, {"word": " learning", "start": 467.84, "end": 468.04, "probability": 0.9939350485801697}, {"word": " frameworks,", "start": 468.04, "end": 468.52, "probability": 0.9932308197021484}]}, {"id": 90, "seek": 45640, "start": 468.78, "end": 474.0, "text": " and so better machine learning models are possible. Federated learning is superior", "tokens": [50975, 290, 523, 1365, 4572, 4673, 4981, 389, 1744, 13, 35089, 515, 4673, 318, 9098, 51243], "temperature": 0.0, "avg_logprob": -0.07136473140201054, "compression_ratio": 1.6009174311926606, "no_speech_prob": 0.002968456130474806, "words": [{"word": " and", "start": 468.78, "end": 468.82, "probability": 0.9692023396492004}, {"word": " so", "start": 468.82, "end": 469.06, "probability": 0.9842531085014343}, {"word": " better", "start": 469.06, "end": 469.46, "probability": 0.9620156288146973}, {"word": " machine", "start": 469.46, "end": 470.24, "probability": 0.9881947040557861}, {"word": " learning", "start": 470.24, "end": 470.48, "probability": 0.9796258807182312}, {"word": " models", "start": 470.48, "end": 470.82, "probability": 0.9984169006347656}, {"word": " are", "start": 470.82, "end": 471.22, "probability": 0.9972113966941833}, {"word": " possible.", "start": 471.22, "end": 471.88, "probability": 0.99817955493927}, {"word": " Federated", "start": 472.76, "end": 472.9, "probability": 0.9792709648609161}, {"word": " learning", "start": 472.9, "end": 473.2, "probability": 0.9757277965545654}, {"word": " is", "start": 473.2, "end": 473.54, "probability": 0.9947972893714905}, {"word": " superior", "start": 473.54, "end": 474.0, "probability": 0.9989853501319885}]}, {"id": 91, "seek": 45640, "start": 474.88, "end": 480.86, "text": " because it uses that data. Organizations want to take your resources, in this case data,", "tokens": [51303, 780, 340, 3544, 326, 1366, 13, 41846, 765, 284, 1011, 534, 4133, 11, 287, 428, 1339, 1366, 11, 51595], "temperature": 0.0, "avg_logprob": -0.07136473140201054, "compression_ratio": 1.6009174311926606, "no_speech_prob": 0.002968456130474806, "words": [{"word": " because", "start": 474.88, "end": 475.56, "probability": 0.7895394563674927}, {"word": " it", "start": 475.56, "end": 475.72, "probability": 0.9985925555229187}, {"word": " uses", "start": 475.72, "end": 475.96, "probability": 0.9990308284759521}, {"word": " that", "start": 475.96, "end": 476.26, "probability": 0.9983295798301697}, {"word": " data.", "start": 476.26, "end": 476.58, "probability": 0.9993550181388855}, {"word": " Organizations", "start": 477.4, "end": 477.4, "probability": 0.968488872051239}, {"word": " want", "start": 477.4, "end": 478.16, "probability": 0.9952294826507568}, {"word": " to", "start": 478.16, "end": 478.38, "probability": 0.9994551539421082}, {"word": " take", "start": 478.38, "end": 478.86, "probability": 0.9990660548210144}, {"word": " your", "start": 478.86, "end": 479.36, "probability": 0.998115062713623}, {"word": " resources,", "start": 479.36, "end": 479.82, "probability": 0.998299777507782}, {"word": " in", "start": 480.12, "end": 480.24, "probability": 0.9971201419830322}, {"word": " this", "start": 480.24, "end": 480.4, "probability": 0.9993732571601868}, {"word": " case", "start": 480.4, "end": 480.56, "probability": 0.9996528625488281}, {"word": " data,", "start": 480.56, "end": 480.86, "probability": 0.6618253588676453}]}, {"id": 92, "seek": 48086, "start": 480.86, "end": 487.42, "text": " and sell you the value, in this case the knowledge stored therein. Free neural networks like", "tokens": [50375, 290, 3677, 345, 262, 1988, 11, 287, 428, 1339, 262, 3725, 8574, 27258, 13, 3232, 17019, 7686, 588, 50667], "temperature": 0.0, "avg_logprob": -0.1865884441218964, "compression_ratio": 1.4352331606217616, "no_speech_prob": 0.003566720988601446, "words": [{"word": " and", "start": 480.86, "end": 481.3, "probability": 0.4883958399295807}, {"word": " sell", "start": 481.3, "end": 481.56, "probability": 0.9273442625999451}, {"word": " you", "start": 481.56, "end": 481.72, "probability": 0.9959677457809448}, {"word": " the", "start": 481.72, "end": 481.84, "probability": 0.9973509311676025}, {"word": " value,", "start": 481.84, "end": 482.14, "probability": 0.9976671934127808}, {"word": " in", "start": 482.28, "end": 482.38, "probability": 0.9859593510627747}, {"word": " this", "start": 482.38, "end": 482.54, "probability": 0.9963000416755676}, {"word": " case", "start": 482.54, "end": 482.76, "probability": 0.9992324113845825}, {"word": " the", "start": 482.76, "end": 483.12, "probability": 0.3618469834327698}, {"word": " knowledge", "start": 483.12, "end": 483.44, "probability": 0.9915066361427307}, {"word": " stored", "start": 483.44, "end": 483.82, "probability": 0.5558121800422668}, {"word": " therein.", "start": 483.82, "end": 484.26, "probability": 0.8308596014976501}, {"word": " Free", "start": 485.44, "end": 485.74, "probability": 0.9812998175621033}, {"word": " neural", "start": 485.74, "end": 485.96, "probability": 0.9515447020530701}, {"word": " networks", "start": 485.96, "end": 486.44, "probability": 0.9980013966560364}, {"word": " like", "start": 486.44, "end": 487.42, "probability": 0.674202024936676}]}, {"id": 93, "seek": 48086, "start": 487.42, "end": 496.24, "text": " LAMA have largely closed the gaps on large language models like chat GPT-3. GPT-4, I know,", "tokens": [50711, 406, 25087, 423, 5688, 4838, 262, 17332, 319, 1588, 3303, 4981, 588, 8537, 402, 11571, 12, 18, 13, 402, 11571, 12, 19, 11, 314, 760, 11, 51131], "temperature": 0.0, "avg_logprob": -0.1865884441218964, "compression_ratio": 1.4352331606217616, "no_speech_prob": 0.003566720988601446, "words": [{"word": " LAMA", "start": 487.42, "end": 488.16, "probability": 0.5644217133522034}, {"word": " have", "start": 488.16, "end": 488.56, "probability": 0.9368937015533447}, {"word": " largely", "start": 488.56, "end": 488.94, "probability": 0.9893447756767273}, {"word": " closed", "start": 488.94, "end": 489.26, "probability": 0.979496955871582}, {"word": " the", "start": 489.26, "end": 489.44, "probability": 0.9973342418670654}, {"word": " gaps", "start": 489.44, "end": 489.72, "probability": 0.997115969657898}, {"word": " on", "start": 489.72, "end": 490.32, "probability": 0.9883593320846558}, {"word": " large", "start": 490.32, "end": 491.86, "probability": 0.3459301292896271}, {"word": " language", "start": 491.86, "end": 492.16, "probability": 0.8525072336196899}, {"word": " models", "start": 492.16, "end": 492.56, "probability": 0.9936356544494629}, {"word": " like", "start": 492.56, "end": 492.78, "probability": 0.9325127005577087}, {"word": " chat", "start": 492.78, "end": 493.02, "probability": 0.3796987235546112}, {"word": " GPT", "start": 493.02, "end": 493.4, "probability": 0.7182013988494873}, {"word": "-3.", "start": 493.4, "end": 493.74, "probability": 0.861809104681015}, {"word": " GPT", "start": 494.74, "end": 495.34, "probability": 0.9911050498485565}, {"word": "-4,", "start": 495.34, "end": 495.8, "probability": 0.9913250803947449}, {"word": " I", "start": 495.82, "end": 496.04, "probability": 0.998375415802002}, {"word": " know,", "start": 496.04, "end": 496.24, "probability": 0.9995360374450684}]}, {"id": 94, "seek": 48086, "start": 496.32, "end": 505.2, "text": " is a leg up, but I do trust that we are making strides and that we will eventually outperform", "tokens": [51131, 318, 257, 1232, 510, 11, 475, 314, 466, 3774, 326, 356, 389, 1642, 35002, 290, 326, 356, 481, 4191, 33597, 687, 51579], "temperature": 0.0, "avg_logprob": -0.1865884441218964, "compression_ratio": 1.4352331606217616, "no_speech_prob": 0.003566720988601446, "words": [{"word": " is", "start": 496.32, "end": 496.74, "probability": 0.9984286427497864}, {"word": " a", "start": 496.74, "end": 496.88, "probability": 0.9792641997337341}, {"word": " leg", "start": 496.88, "end": 497.06, "probability": 0.9899987578392029}, {"word": " up,", "start": 497.06, "end": 497.3, "probability": 0.9438273310661316}, {"word": " but", "start": 497.52, "end": 497.64, "probability": 0.9985547661781311}, {"word": " I", "start": 497.64, "end": 497.96, "probability": 0.9980065226554871}, {"word": " do", "start": 497.96, "end": 498.28, "probability": 0.9987543821334839}, {"word": " trust", "start": 498.28, "end": 498.62, "probability": 0.9992766976356506}, {"word": " that", "start": 498.62, "end": 499.32, "probability": 0.9966009259223938}, {"word": " we", "start": 499.32, "end": 501.62, "probability": 0.9349637627601624}, {"word": " are", "start": 501.62, "end": 501.98, "probability": 0.9966654181480408}, {"word": " making", "start": 501.98, "end": 502.28, "probability": 0.9995730519294739}, {"word": " strides", "start": 502.28, "end": 502.68, "probability": 0.9987049102783203}, {"word": " and", "start": 502.68, "end": 503.2, "probability": 0.8761547803878784}, {"word": " that", "start": 503.2, "end": 503.48, "probability": 0.9901682138442993}, {"word": " we", "start": 503.48, "end": 503.7, "probability": 0.9975020289421082}, {"word": " will", "start": 503.7, "end": 504.02, "probability": 0.9984641075134277}, {"word": " eventually", "start": 504.02, "end": 504.56, "probability": 0.9934473633766174}, {"word": " outperform", "start": 504.56, "end": 505.2, "probability": 0.9678997993469238}]}, {"id": 95, "seek": 50520, "start": 505.2, "end": 512.24, "text": " them simply because there's more data. We don't need centralized AI gatekeepers to control AI and", "tokens": [50363, 606, 2391, 780, 612, 338, 517, 1366, 13, 775, 836, 470, 761, 29024, 9552, 8946, 24952, 284, 1630, 9552, 290, 50715], "temperature": 0.0, "avg_logprob": -0.13875909047584012, "compression_ratio": 1.4558823529411764, "no_speech_prob": 0.1373765468597412, "words": [{"word": " them", "start": 505.2, "end": 505.38, "probability": 0.376903235912323}, {"word": " simply", "start": 505.38, "end": 505.8, "probability": 0.8722171783447266}, {"word": " because", "start": 505.8, "end": 506.22, "probability": 0.9971855282783508}, {"word": " there's", "start": 506.22, "end": 506.52, "probability": 0.9506147503852844}, {"word": " more", "start": 506.52, "end": 506.72, "probability": 0.9983901977539062}, {"word": " data.", "start": 506.72, "end": 507.02, "probability": 0.995955228805542}, {"word": " We", "start": 508.14, "end": 508.32, "probability": 0.9175779819488525}, {"word": " don't", "start": 508.32, "end": 508.46, "probability": 0.9983539283275604}, {"word": " need", "start": 508.46, "end": 508.72, "probability": 0.9986962676048279}, {"word": " centralized", "start": 508.72, "end": 509.7, "probability": 0.8844448328018188}, {"word": " AI", "start": 509.7, "end": 510.18, "probability": 0.8509461879730225}, {"word": " gatekeepers", "start": 510.18, "end": 511.0, "probability": 0.9884056746959686}, {"word": " to", "start": 511.0, "end": 511.24, "probability": 0.9965658783912659}, {"word": " control", "start": 511.24, "end": 511.66, "probability": 0.9963501691818237}, {"word": " AI", "start": 511.66, "end": 511.9, "probability": 0.9912079572677612}, {"word": " and", "start": 511.9, "end": 512.24, "probability": 0.9838990569114685}]}, {"id": 96, "seek": 50520, "start": 512.24, "end": 521.18, "text": " protect us from ourselves. I'll end my talk with an updated subtitle, and here I mean free as in", "tokens": [50715, 1805, 514, 422, 6731, 13, 314, 1183, 886, 616, 1561, 351, 281, 6153, 37960, 11, 290, 994, 314, 1612, 1479, 355, 287, 51159], "temperature": 0.0, "avg_logprob": -0.13875909047584012, "compression_ratio": 1.4558823529411764, "no_speech_prob": 0.1373765468597412, "words": [{"word": " protect", "start": 512.24, "end": 512.62, "probability": 0.9986781477928162}, {"word": " us", "start": 512.62, "end": 512.94, "probability": 0.9936969876289368}, {"word": " from", "start": 512.94, "end": 513.22, "probability": 0.9976518750190735}, {"word": " ourselves.", "start": 513.22, "end": 513.66, "probability": 0.997306227684021}, {"word": " I'll", "start": 514.94, "end": 515.62, "probability": 0.6283546537160873}, {"word": " end", "start": 515.62, "end": 515.9, "probability": 0.9990748167037964}, {"word": " my", "start": 515.9, "end": 516.16, "probability": 0.9964588284492493}, {"word": " talk", "start": 516.16, "end": 516.52, "probability": 0.9987429976463318}, {"word": " with", "start": 516.52, "end": 517.14, "probability": 0.9979943037033081}, {"word": " an", "start": 517.14, "end": 517.56, "probability": 0.9929034113883972}, {"word": " updated", "start": 517.56, "end": 518.06, "probability": 0.9971814155578613}, {"word": " subtitle,", "start": 518.06, "end": 518.62, "probability": 0.9794737696647644}, {"word": " and", "start": 519.22, "end": 519.4, "probability": 0.984100878238678}, {"word": " here", "start": 519.4, "end": 519.9, "probability": 0.9862663745880127}, {"word": " I", "start": 519.9, "end": 520.26, "probability": 0.8392001986503601}, {"word": " mean", "start": 520.26, "end": 520.56, "probability": 0.9962074756622314}, {"word": " free", "start": 520.56, "end": 520.88, "probability": 0.6557295918464661}, {"word": " as", "start": 520.88, "end": 521.04, "probability": 0.4652356505393982}, {"word": " in", "start": 521.04, "end": 521.18, "probability": 0.9732228517532349}]}, {"id": 97, "seek": 50520, "start": 521.18, "end": 533.78, "text": " freedom. Avoid open AI, embrace free AI. Thank you. Any questions? We've got one minute for questions.", "tokens": [51159, 4925, 13, 24390, 1280, 9552, 11, 12553, 1479, 9552, 13, 6952, 345, 13, 4377, 2683, 30, 775, 1053, 1392, 530, 5664, 329, 2683, 13, 51795], "temperature": 0.0, "avg_logprob": -0.13875909047584012, "compression_ratio": 1.4558823529411764, "no_speech_prob": 0.1373765468597412, "words": [{"word": " freedom.", "start": 521.18, "end": 521.48, "probability": 0.9923678040504456}, {"word": " Avoid", "start": 522.5, "end": 522.62, "probability": 0.9048153162002563}, {"word": " open", "start": 522.62, "end": 523.02, "probability": 0.6797588467597961}, {"word": " AI,", "start": 523.02, "end": 523.4, "probability": 0.9866840243339539}, {"word": " embrace", "start": 523.86, "end": 524.14, "probability": 0.8789015412330627}, {"word": " free", "start": 524.14, "end": 524.74, "probability": 0.9954857230186462}, {"word": " AI.", "start": 524.74, "end": 525.32, "probability": 0.9981841444969177}, {"word": " Thank", "start": 527.42, "end": 527.64, "probability": 0.9199866056442261}, {"word": " you.", "start": 527.64, "end": 527.86, "probability": 0.9975795149803162}, {"word": " Any", "start": 531.5, "end": 531.8, "probability": 0.9178212285041809}, {"word": " questions?", "start": 531.8, "end": 532.1, "probability": 0.9963144659996033}, {"word": " We've", "start": 532.6, "end": 532.78, "probability": 0.8460507690906525}, {"word": " got", "start": 532.78, "end": 532.88, "probability": 0.9938830137252808}, {"word": " one", "start": 532.88, "end": 533.04, "probability": 0.9800772666931152}, {"word": " minute", "start": 533.04, "end": 533.2, "probability": 0.8851361870765686}, {"word": " for", "start": 533.2, "end": 533.4, "probability": 0.9852136969566345}, {"word": " questions.", "start": 533.4, "end": 533.78, "probability": 0.9479631185531616}]}, {"id": 98, "seek": 53520, "start": 535.2, "end": 544.8, "text": " I have a question. So the kind of framing of the threat model here is companies want to take your", "tokens": [50363, 314, 423, 257, 1808, 13, 1406, 262, 1611, 286, 30811, 286, 262, 2372, 2746, 994, 318, 2706, 765, 284, 1011, 534, 50839], "temperature": 0.0, "avg_logprob": -0.12844183111703525, "compression_ratio": 1.695852534562212, "no_speech_prob": 0.027558661997318268, "words": [{"word": " I", "start": 535.2, "end": 535.2, "probability": 0.13593950867652893}, {"word": " have", "start": 535.2, "end": 535.26, "probability": 0.9422591328620911}, {"word": " a", "start": 535.26, "end": 535.36, "probability": 0.9918477535247803}, {"word": " question.", "start": 535.36, "end": 535.68, "probability": 0.9967973828315735}, {"word": " So", "start": 536.26, "end": 536.26, "probability": 0.645248532295227}, {"word": " the", "start": 536.26, "end": 537.96, "probability": 0.3433716297149658}, {"word": " kind", "start": 537.96, "end": 538.7, "probability": 0.6844940185546875}, {"word": " of", "start": 538.7, "end": 539.2, "probability": 0.9934386610984802}, {"word": " framing", "start": 539.2, "end": 539.98, "probability": 0.9908275008201599}, {"word": " of", "start": 539.98, "end": 540.2, "probability": 0.9927910566329956}, {"word": " the", "start": 540.2, "end": 540.32, "probability": 0.9763452410697937}, {"word": " threat", "start": 540.32, "end": 540.52, "probability": 0.9711704254150391}, {"word": " model", "start": 540.52, "end": 540.86, "probability": 0.9954400062561035}, {"word": " here", "start": 540.86, "end": 541.2, "probability": 0.9869911074638367}, {"word": " is", "start": 541.2, "end": 541.86, "probability": 0.9661376476287842}, {"word": " companies", "start": 541.86, "end": 543.78, "probability": 0.7093309760093689}, {"word": " want", "start": 543.78, "end": 544.14, "probability": 0.9710965752601624}, {"word": " to", "start": 544.14, "end": 544.42, "probability": 0.9936902523040771}, {"word": " take", "start": 544.42, "end": 544.64, "probability": 0.7909227609634399}, {"word": " your", "start": 544.64, "end": 544.8, "probability": 0.5094845294952393}]}, {"id": 99, "seek": 53520, "start": 544.8, "end": 550.76, "text": " data so that they can train models off of it and sell it back to you. How does federated learning", "tokens": [50839, 1366, 523, 326, 484, 460, 4512, 4981, 572, 286, 340, 290, 3677, 340, 736, 284, 345, 13, 1374, 857, 28062, 515, 4673, 51139], "temperature": 0.0, "avg_logprob": -0.12844183111703525, "compression_ratio": 1.695852534562212, "no_speech_prob": 0.027558661997318268, "words": [{"word": " data", "start": 544.8, "end": 544.98, "probability": 0.9975175857543945}, {"word": " so", "start": 544.98, "end": 545.26, "probability": 0.9624069929122925}, {"word": " that", "start": 545.26, "end": 545.4, "probability": 0.8891516923904419}, {"word": " they", "start": 545.4, "end": 545.52, "probability": 0.9985998272895813}, {"word": " can", "start": 545.52, "end": 545.68, "probability": 0.9984729886054993}, {"word": " train", "start": 545.68, "end": 545.9, "probability": 0.9870868921279907}, {"word": " models", "start": 545.9, "end": 546.22, "probability": 0.988860547542572}, {"word": " off", "start": 546.22, "end": 546.46, "probability": 0.7921580672264099}, {"word": " of", "start": 546.46, "end": 546.62, "probability": 0.9883759021759033}, {"word": " it", "start": 546.62, "end": 546.86, "probability": 0.9982531666755676}, {"word": " and", "start": 546.86, "end": 547.36, "probability": 0.9260330200195312}, {"word": " sell", "start": 547.36, "end": 548.28, "probability": 0.5613781213760376}, {"word": " it", "start": 548.28, "end": 548.38, "probability": 0.9580639004707336}, {"word": " back", "start": 548.38, "end": 548.54, "probability": 0.9923093318939209}, {"word": " to", "start": 548.54, "end": 548.74, "probability": 0.9937794804573059}, {"word": " you.", "start": 548.74, "end": 549.0, "probability": 0.9972884654998779}, {"word": " How", "start": 549.42, "end": 549.62, "probability": 0.9911544322967529}, {"word": " does", "start": 549.62, "end": 549.9, "probability": 0.9924336671829224}, {"word": " federated", "start": 549.9, "end": 550.32, "probability": 0.9232643842697144}, {"word": " learning", "start": 550.32, "end": 550.76, "probability": 0.9980891346931458}]}, {"id": 100, "seek": 53520, "start": 551.3599999999999, "end": 555.68, "text": " change that? Now they don't need to take the data, but they still are training a model", "tokens": [51179, 1487, 326, 30, 2735, 484, 836, 470, 761, 284, 1011, 262, 1366, 11, 475, 484, 991, 389, 3047, 257, 2746, 51383], "temperature": 0.0, "avg_logprob": -0.12844183111703525, "compression_ratio": 1.695852534562212, "no_speech_prob": 0.027558661997318268, "words": [{"word": " change", "start": 551.3599999999999, "end": 551.8, "probability": 0.9942523837089539}, {"word": " that?", "start": 551.8, "end": 552.1, "probability": 0.9922841191291809}, {"word": " Now", "start": 552.58, "end": 553.02, "probability": 0.796461284160614}, {"word": " they", "start": 553.02, "end": 553.16, "probability": 0.8838736414909363}, {"word": " don't", "start": 553.16, "end": 553.28, "probability": 0.9951589107513428}, {"word": " need", "start": 553.28, "end": 553.42, "probability": 0.8825423121452332}, {"word": " to", "start": 553.42, "end": 553.5, "probability": 0.9957513809204102}, {"word": " take", "start": 553.5, "end": 553.7, "probability": 0.9965554475784302}, {"word": " the", "start": 553.7, "end": 553.86, "probability": 0.9872021079063416}, {"word": " data,", "start": 553.86, "end": 554.08, "probability": 0.999627947807312}, {"word": " but", "start": 554.16, "end": 554.26, "probability": 0.9977044463157654}, {"word": " they", "start": 554.26, "end": 554.4, "probability": 0.9974727034568787}, {"word": " still", "start": 554.4, "end": 554.62, "probability": 0.943743109703064}, {"word": " are", "start": 554.62, "end": 554.76, "probability": 0.9902530908584595}, {"word": " training", "start": 554.76, "end": 555.18, "probability": 0.9945371747016907}, {"word": " a", "start": 555.18, "end": 555.36, "probability": 0.981826663017273}, {"word": " model", "start": 555.36, "end": 555.68, "probability": 0.999528169631958}]}, {"id": 101, "seek": 53520, "start": 555.68, "end": 560.06, "text": " without me that they can sell back to me. But they're not training it without you. So", "tokens": [51411, 1231, 502, 326, 484, 460, 3677, 736, 284, 502, 13, 887, 484, 821, 407, 3047, 340, 1231, 345, 13, 1406, 51595], "temperature": 0.0, "avg_logprob": -0.12844183111703525, "compression_ratio": 1.695852534562212, "no_speech_prob": 0.027558661997318268, "words": [{"word": " without", "start": 555.68, "end": 556.54, "probability": 0.9819891452789307}, {"word": " me", "start": 556.54, "end": 556.74, "probability": 0.9884200096130371}, {"word": " that", "start": 556.74, "end": 556.88, "probability": 0.9443472027778625}, {"word": " they", "start": 556.88, "end": 557.04, "probability": 0.9982693195343018}, {"word": " can", "start": 557.04, "end": 557.16, "probability": 0.9843952655792236}, {"word": " sell", "start": 557.16, "end": 557.34, "probability": 0.9964354038238525}, {"word": " back", "start": 557.34, "end": 557.56, "probability": 0.9104517102241516}, {"word": " to", "start": 557.56, "end": 557.74, "probability": 0.9822109341621399}, {"word": " me.", "start": 557.74, "end": 557.9, "probability": 0.9853878617286682}, {"word": " But", "start": 558.04, "end": 558.2, "probability": 0.6047173142433167}, {"word": " they're", "start": 558.2, "end": 558.44, "probability": 0.9866129755973816}, {"word": " not", "start": 558.44, "end": 558.6, "probability": 0.9993725419044495}, {"word": " training", "start": 558.6, "end": 558.88, "probability": 0.9686084389686584}, {"word": " it", "start": 558.88, "end": 559.06, "probability": 0.9872194528579712}, {"word": " without", "start": 559.06, "end": 559.28, "probability": 0.999150276184082}, {"word": " you.", "start": 559.28, "end": 559.54, "probability": 0.9991742968559265}, {"word": " So", "start": 559.88, "end": 560.06, "probability": 0.9611465334892273}]}, {"id": 102, "seek": 56006, "start": 560.6199999999999, "end": 567.26, "text": " every node has to have access to the model weights in order for you to converge on one set of model", "tokens": [50391, 790, 10139, 468, 284, 423, 1895, 284, 262, 2746, 19590, 287, 1502, 329, 345, 284, 47873, 319, 530, 900, 286, 2746, 50723], "temperature": 0.0, "avg_logprob": -0.0846926734561012, "compression_ratio": 1.6694214876033058, "no_speech_prob": 0.013591751456260681, "words": [{"word": " every", "start": 560.6199999999999, "end": 560.98, "probability": 0.06686792522668839}, {"word": " node", "start": 560.98, "end": 561.38, "probability": 0.9330092668533325}, {"word": " has", "start": 561.38, "end": 561.66, "probability": 0.9783671498298645}, {"word": " to", "start": 561.66, "end": 561.78, "probability": 0.9863409399986267}, {"word": " have", "start": 561.78, "end": 561.94, "probability": 0.9949949979782104}, {"word": " access", "start": 561.94, "end": 562.4, "probability": 0.9787477254867554}, {"word": " to", "start": 562.4, "end": 562.9, "probability": 0.9902647733688354}, {"word": " the", "start": 562.9, "end": 563.16, "probability": 0.9856902956962585}, {"word": " model", "start": 563.16, "end": 563.48, "probability": 0.9826735854148865}, {"word": " weights", "start": 563.48, "end": 563.8, "probability": 0.9889912605285645}, {"word": " in", "start": 563.8, "end": 564.42, "probability": 0.869503915309906}, {"word": " order", "start": 564.42, "end": 564.7, "probability": 0.9870720505714417}, {"word": " for", "start": 564.7, "end": 565.18, "probability": 0.9892983436584473}, {"word": " you", "start": 565.18, "end": 565.44, "probability": 0.9971610307693481}, {"word": " to", "start": 565.44, "end": 565.66, "probability": 0.9973799586296082}, {"word": " converge", "start": 565.66, "end": 566.08, "probability": 0.9047955870628357}, {"word": " on", "start": 566.08, "end": 566.5, "probability": 0.9888594150543213}, {"word": " one", "start": 566.5, "end": 566.72, "probability": 0.9883762001991272}, {"word": " set", "start": 566.72, "end": 566.92, "probability": 0.9914796352386475}, {"word": " of", "start": 566.92, "end": 567.06, "probability": 0.9976534247398376}, {"word": " model", "start": 567.06, "end": 567.26, "probability": 0.9925665855407715}]}, {"id": 103, "seek": 56006, "start": 567.26, "end": 573.98, "text": " weights. So the community owns the model together. So you're not being sold it. You put in your work.", "tokens": [50723, 19590, 13, 1406, 262, 2055, 12216, 262, 2746, 1978, 13, 1406, 345, 821, 407, 852, 2702, 340, 13, 921, 1234, 287, 534, 670, 13, 51059], "temperature": 0.0, "avg_logprob": -0.0846926734561012, "compression_ratio": 1.6694214876033058, "no_speech_prob": 0.013591751456260681, "words": [{"word": " weights.", "start": 567.26, "end": 567.52, "probability": 0.9940078258514404}, {"word": " So", "start": 567.94, "end": 567.94, "probability": 0.8998593688011169}, {"word": " the", "start": 567.94, "end": 568.08, "probability": 0.8841397166252136}, {"word": " community", "start": 568.08, "end": 568.48, "probability": 0.9869426488876343}, {"word": " owns", "start": 568.48, "end": 569.2, "probability": 0.9954008460044861}, {"word": " the", "start": 569.2, "end": 569.46, "probability": 0.9958555102348328}, {"word": " model", "start": 569.46, "end": 569.76, "probability": 0.9986224174499512}, {"word": " together.", "start": 569.76, "end": 570.26, "probability": 0.9961373209953308}, {"word": " So", "start": 570.8, "end": 571.32, "probability": 0.9666493535041809}, {"word": " you're", "start": 571.32, "end": 571.6, "probability": 0.9646917581558228}, {"word": " not", "start": 571.6, "end": 571.74, "probability": 0.9995418787002563}, {"word": " being", "start": 571.74, "end": 571.94, "probability": 0.9974954724311829}, {"word": " sold", "start": 571.94, "end": 572.24, "probability": 0.9967151880264282}, {"word": " it.", "start": 572.24, "end": 572.5, "probability": 0.9476882815361023}, {"word": " You", "start": 572.82, "end": 573.04, "probability": 0.9902451038360596}, {"word": " put", "start": 573.04, "end": 573.32, "probability": 0.9456617832183838}, {"word": " in", "start": 573.32, "end": 573.48, "probability": 0.9972519278526306}, {"word": " your", "start": 573.48, "end": 573.66, "probability": 0.9977454543113708}, {"word": " work.", "start": 573.66, "end": 573.98, "probability": 0.9984851479530334}]}, {"id": 104, "seek": 56006, "start": 574.16, "end": 580.82, "text": " You said, I have this data. I want valuable tools. Maybe it's like I have a voice recognizer. I have", "tokens": [51059, 921, 531, 11, 314, 423, 428, 1366, 13, 314, 765, 8119, 4899, 13, 6674, 340, 338, 588, 314, 423, 257, 3809, 3018, 7509, 13, 314, 423, 51399], "temperature": 0.0, "avg_logprob": -0.0846926734561012, "compression_ratio": 1.6694214876033058, "no_speech_prob": 0.013591751456260681, "words": [{"word": " You", "start": 574.16, "end": 574.2, "probability": 0.9938750863075256}, {"word": " said,", "start": 574.2, "end": 574.4, "probability": 0.9964044094085693}, {"word": " I", "start": 574.5, "end": 574.62, "probability": 0.8833327889442444}, {"word": " have", "start": 574.62, "end": 574.78, "probability": 0.9947460889816284}, {"word": " this", "start": 574.78, "end": 574.94, "probability": 0.9990589022636414}, {"word": " data.", "start": 574.94, "end": 575.24, "probability": 0.9987927675247192}, {"word": " I", "start": 575.76, "end": 575.8, "probability": 0.998672366142273}, {"word": " want", "start": 575.8, "end": 576.0, "probability": 0.9984083771705627}, {"word": " valuable", "start": 576.0, "end": 576.44, "probability": 0.9972808361053467}, {"word": " tools.", "start": 576.44, "end": 577.12, "probability": 0.9977273344993591}, {"word": " Maybe", "start": 577.78, "end": 577.9, "probability": 0.7990079522132874}, {"word": " it's", "start": 577.9, "end": 578.16, "probability": 0.9706577062606812}, {"word": " like", "start": 578.16, "end": 578.64, "probability": 0.6537867188453674}, {"word": " I", "start": 578.64, "end": 578.8, "probability": 0.5934987664222717}, {"word": " have", "start": 578.8, "end": 578.9, "probability": 0.9930992126464844}, {"word": " a", "start": 578.9, "end": 579.0, "probability": 0.9988657236099243}, {"word": " voice", "start": 579.0, "end": 579.2, "probability": 0.9947463870048523}, {"word": " recognizer.", "start": 579.2, "end": 579.82, "probability": 0.9563280642032623}, {"word": " I", "start": 580.02, "end": 580.28, "probability": 0.995384156703949}, {"word": " have", "start": 580.28, "end": 580.82, "probability": 0.8045719265937805}]}, {"id": 105, "seek": 56006, "start": 580.82, "end": 586.52, "text": " a different accent, so I need to add some data for my accent so that it can work. And so I contribute", "tokens": [51399, 257, 1180, 18702, 11, 523, 314, 761, 284, 751, 617, 1366, 329, 616, 18702, 523, 326, 340, 460, 670, 13, 843, 523, 314, 8676, 51687], "temperature": 0.0, "avg_logprob": -0.0846926734561012, "compression_ratio": 1.6694214876033058, "no_speech_prob": 0.013591751456260681, "words": [{"word": " a", "start": 580.82, "end": 580.94, "probability": 0.9986827969551086}, {"word": " different", "start": 580.94, "end": 581.14, "probability": 0.9996111989021301}, {"word": " accent,", "start": 581.14, "end": 581.52, "probability": 0.9677468538284302}, {"word": " so", "start": 581.68, "end": 581.72, "probability": 0.9887054562568665}, {"word": " I", "start": 581.72, "end": 581.84, "probability": 0.9975906610488892}, {"word": " need", "start": 581.84, "end": 581.96, "probability": 0.9983795881271362}, {"word": " to", "start": 581.96, "end": 582.16, "probability": 0.9984443783760071}, {"word": " add", "start": 582.16, "end": 582.58, "probability": 0.9955119490623474}, {"word": " some", "start": 582.58, "end": 582.84, "probability": 0.9982045888900757}, {"word": " data", "start": 582.84, "end": 583.18, "probability": 0.9979304075241089}, {"word": " for", "start": 583.18, "end": 583.48, "probability": 0.9888706803321838}, {"word": " my", "start": 583.48, "end": 583.7, "probability": 0.9988775849342346}, {"word": " accent", "start": 583.7, "end": 584.1, "probability": 0.9838835597038269}, {"word": " so", "start": 584.1, "end": 584.38, "probability": 0.968723714351654}, {"word": " that", "start": 584.38, "end": 584.56, "probability": 0.9992606043815613}, {"word": " it", "start": 584.56, "end": 584.68, "probability": 0.9986131191253662}, {"word": " can", "start": 584.68, "end": 584.88, "probability": 0.9988619089126587}, {"word": " work.", "start": 584.88, "end": 585.24, "probability": 0.9977776408195496}, {"word": " And", "start": 585.7, "end": 585.76, "probability": 0.9734438061714172}, {"word": " so", "start": 585.76, "end": 585.98, "probability": 0.9957036375999451}, {"word": " I", "start": 585.98, "end": 586.14, "probability": 0.9848828911781311}, {"word": " contribute", "start": 586.14, "end": 586.52, "probability": 0.9911128878593445}]}, {"id": 106, "seek": 58652, "start": 586.52, "end": 592.36, "text": " to learn the shared model and I get something out of it and the overall global model benefits as well.", "tokens": [50363, 284, 2193, 262, 4888, 2746, 290, 314, 651, 1223, 503, 286, 340, 290, 262, 4045, 3298, 2746, 4034, 355, 880, 13, 50655], "temperature": 0.0, "avg_logprob": -0.1731940200648357, "compression_ratio": 1.6565217391304348, "no_speech_prob": 0.015098645351827145, "words": [{"word": " to", "start": 586.52, "end": 586.72, "probability": 0.1043330654501915}, {"word": " learn", "start": 586.72, "end": 586.88, "probability": 0.8644320964813232}, {"word": " the", "start": 586.88, "end": 587.04, "probability": 0.9798254370689392}, {"word": " shared", "start": 587.04, "end": 587.22, "probability": 0.9590244293212891}, {"word": " model", "start": 587.22, "end": 587.5, "probability": 0.9922237396240234}, {"word": " and", "start": 587.5, "end": 587.66, "probability": 0.7294363975524902}, {"word": " I", "start": 587.66, "end": 587.8, "probability": 0.9134894013404846}, {"word": " get", "start": 587.8, "end": 587.92, "probability": 0.995278000831604}, {"word": " something", "start": 587.92, "end": 588.16, "probability": 0.9946488738059998}, {"word": " out", "start": 588.16, "end": 588.4, "probability": 0.9980126619338989}, {"word": " of", "start": 588.4, "end": 588.54, "probability": 0.9979350566864014}, {"word": " it", "start": 588.54, "end": 588.7, "probability": 0.9982132911682129}, {"word": " and", "start": 588.7, "end": 589.08, "probability": 0.5841473937034607}, {"word": " the", "start": 589.08, "end": 589.22, "probability": 0.9964703321456909}, {"word": " overall", "start": 589.22, "end": 589.58, "probability": 0.9967543482780457}, {"word": " global", "start": 589.58, "end": 589.96, "probability": 0.9914801120758057}, {"word": " model", "start": 589.96, "end": 590.52, "probability": 0.996656060218811}, {"word": " benefits", "start": 590.52, "end": 591.88, "probability": 0.9225585460662842}, {"word": " as", "start": 591.88, "end": 592.16, "probability": 0.9740480780601501}, {"word": " well.", "start": 592.16, "end": 592.4, "probability": 0.9982016086578369}]}, {"id": 107, "seek": 58652, "start": 593.24, "end": 599.94, "text": " So when it's federated, you are always getting a copy of the model locally. Right. Right.", "tokens": [50687, 1406, 618, 340, 338, 28062, 515, 11, 345, 389, 1464, 1972, 257, 4866, 286, 262, 2746, 15726, 13, 6498, 13, 6498, 13, 51031], "temperature": 0.0, "avg_logprob": -0.1731940200648357, "compression_ratio": 1.6565217391304348, "no_speech_prob": 0.015098645351827145, "words": [{"word": " So", "start": 593.24, "end": 593.26, "probability": 0.6468885540962219}, {"word": " when", "start": 593.26, "end": 593.42, "probability": 0.7943034768104553}, {"word": " it's", "start": 593.42, "end": 595.04, "probability": 0.9668291509151459}, {"word": " federated,", "start": 595.04, "end": 595.4, "probability": 0.5619084388017654}, {"word": " you", "start": 595.62, "end": 595.62, "probability": 0.9869555830955505}, {"word": " are", "start": 595.62, "end": 595.8, "probability": 0.9369955062866211}, {"word": " always", "start": 595.8, "end": 596.24, "probability": 0.9965039491653442}, {"word": " getting", "start": 596.24, "end": 596.68, "probability": 0.9727646708488464}, {"word": " a", "start": 596.68, "end": 596.9, "probability": 0.9963477253913879}, {"word": " copy", "start": 596.9, "end": 597.14, "probability": 0.9989423155784607}, {"word": " of", "start": 597.14, "end": 597.36, "probability": 0.9961436986923218}, {"word": " the", "start": 597.36, "end": 597.76, "probability": 0.9899277091026306}, {"word": " model", "start": 597.76, "end": 598.28, "probability": 0.9955350160598755}, {"word": " locally.", "start": 598.28, "end": 598.76, "probability": 0.9786560535430908}, {"word": " Right.", "start": 599.18, "end": 599.32, "probability": 0.793099582195282}, {"word": " Right.", "start": 599.94, "end": 599.94, "probability": 0.634979248046875}]}, {"id": 108, "seek": 58652, "start": 601.14, "end": 607.48, "text": " At that moment in time. Yes. That's the whole training loop is you get a model parameters.", "tokens": [51099, 1629, 326, 2589, 287, 640, 13, 3363, 13, 1320, 338, 262, 2187, 3047, 9052, 318, 345, 651, 257, 2746, 10007, 13, 51407], "temperature": 0.0, "avg_logprob": -0.1731940200648357, "compression_ratio": 1.6565217391304348, "no_speech_prob": 0.015098645351827145, "words": [{"word": " At", "start": 601.14, "end": 601.28, "probability": 0.8326955437660217}, {"word": " that", "start": 601.28, "end": 601.66, "probability": 0.9875875115394592}, {"word": " moment", "start": 601.66, "end": 601.88, "probability": 0.9666317105293274}, {"word": " in", "start": 601.88, "end": 602.08, "probability": 0.8376454710960388}, {"word": " time.", "start": 602.08, "end": 602.34, "probability": 0.9968685507774353}, {"word": " Yes.", "start": 602.56, "end": 602.76, "probability": 0.9407455325126648}, {"word": " That's", "start": 603.58, "end": 603.88, "probability": 0.9095858335494995}, {"word": " the", "start": 603.88, "end": 604.52, "probability": 0.7933685183525085}, {"word": " whole", "start": 604.52, "end": 604.74, "probability": 0.9967790246009827}, {"word": " training", "start": 604.74, "end": 605.1, "probability": 0.997934103012085}, {"word": " loop", "start": 605.1, "end": 605.34, "probability": 0.9387508630752563}, {"word": " is", "start": 605.34, "end": 605.62, "probability": 0.601198673248291}, {"word": " you", "start": 605.62, "end": 606.02, "probability": 0.8176290988922119}, {"word": " get", "start": 606.02, "end": 606.36, "probability": 0.9392569661140442}, {"word": " a", "start": 606.36, "end": 606.54, "probability": 0.8636906147003174}, {"word": " model", "start": 606.54, "end": 607.08, "probability": 0.9924881458282471}, {"word": " parameters.", "start": 607.08, "end": 607.48, "probability": 0.9653409123420715}]}, {"id": 109, "seek": 58652, "start": 608.86, "end": 614.24, "text": " As far as we know, it's the best model weights or parameters, and you tune them a little bit more", "tokens": [51503, 1081, 1290, 355, 356, 760, 11, 340, 338, 262, 1266, 2746, 19590, 393, 10007, 11, 290, 345, 14009, 606, 257, 1310, 1643, 517, 51747], "temperature": 0.0, "avg_logprob": -0.1731940200648357, "compression_ratio": 1.6565217391304348, "no_speech_prob": 0.015098645351827145, "words": [{"word": " As", "start": 608.86, "end": 609.34, "probability": 0.12993744015693665}, {"word": " far", "start": 609.34, "end": 609.82, "probability": 0.9804733991622925}, {"word": " as", "start": 609.82, "end": 609.92, "probability": 0.9984543323516846}, {"word": " we", "start": 609.92, "end": 610.02, "probability": 0.988264799118042}, {"word": " know,", "start": 610.02, "end": 610.14, "probability": 0.9974143505096436}, {"word": " it's", "start": 610.22, "end": 610.22, "probability": 0.9610257744789124}, {"word": " the", "start": 610.22, "end": 610.32, "probability": 0.9965031147003174}, {"word": " best", "start": 610.32, "end": 610.5, "probability": 0.9984204769134521}, {"word": " model", "start": 610.5, "end": 610.76, "probability": 0.9933093786239624}, {"word": " weights", "start": 610.76, "end": 611.3, "probability": 0.930802583694458}, {"word": " or", "start": 611.3, "end": 612.16, "probability": 0.5011815428733826}, {"word": " parameters,", "start": 612.16, "end": 612.5, "probability": 0.9908804893493652}, {"word": " and", "start": 612.74, "end": 613.0, "probability": 0.9915004968643188}, {"word": " you", "start": 613.0, "end": 613.16, "probability": 0.9980606436729431}, {"word": " tune", "start": 613.16, "end": 613.42, "probability": 0.9826222658157349}, {"word": " them", "start": 613.42, "end": 613.58, "probability": 0.9875183701515198}, {"word": " a", "start": 613.58, "end": 613.7, "probability": 0.9971650242805481}, {"word": " little", "start": 613.7, "end": 613.86, "probability": 0.9996871948242188}, {"word": " bit", "start": 613.86, "end": 614.02, "probability": 0.9971073269844055}, {"word": " more", "start": 614.02, "end": 614.24, "probability": 0.9989782571792603}]}, {"id": 110, "seek": 61424, "start": 614.24, "end": 617.58, "text": " to improve them just a little bit. And then you send those back out into the world and you're", "tokens": [50363, 284, 2987, 606, 655, 257, 1310, 1643, 13, 843, 788, 345, 3758, 883, 736, 503, 656, 262, 995, 290, 345, 821, 50523], "temperature": 0.0, "avg_logprob": -0.15742188579631303, "compression_ratio": 1.6900369003690037, "no_speech_prob": 0.0047852229326963425, "words": [{"word": " to", "start": 614.24, "end": 614.4, "probability": 0.10721893608570099}, {"word": " improve", "start": 614.4, "end": 614.7, "probability": 0.9028543829917908}, {"word": " them", "start": 614.7, "end": 614.9, "probability": 0.9479526281356812}, {"word": " just", "start": 614.9, "end": 615.06, "probability": 0.890679657459259}, {"word": " a", "start": 615.06, "end": 615.2, "probability": 0.9963575005531311}, {"word": " little", "start": 615.2, "end": 615.34, "probability": 0.9879647493362427}, {"word": " bit.", "start": 615.34, "end": 615.58, "probability": 0.9900339245796204}, {"word": " And", "start": 615.88, "end": 615.9, "probability": 0.8535702228546143}, {"word": " then", "start": 615.9, "end": 616.0, "probability": 0.9780527949333191}, {"word": " you", "start": 616.0, "end": 616.14, "probability": 0.9535239338874817}, {"word": " send", "start": 616.14, "end": 616.3, "probability": 0.9903340935707092}, {"word": " those", "start": 616.3, "end": 616.44, "probability": 0.9793944954872131}, {"word": " back", "start": 616.44, "end": 616.66, "probability": 0.9955136179924011}, {"word": " out", "start": 616.66, "end": 616.82, "probability": 0.9734473824501038}, {"word": " into", "start": 616.82, "end": 616.94, "probability": 0.9410067796707153}, {"word": " the", "start": 616.94, "end": 617.04, "probability": 0.9988762736320496}, {"word": " world", "start": 617.04, "end": 617.2, "probability": 0.9983933568000793}, {"word": " and", "start": 617.2, "end": 617.36, "probability": 0.32701513171195984}, {"word": " you're", "start": 617.36, "end": 617.58, "probability": 0.981785923242569}]}, {"id": 111, "seek": 61424, "start": 617.58, "end": 622.4, "text": " constantly doing that. And if you do that enough, you converge to a better model because there's", "tokens": [50523, 7558, 1804, 326, 13, 843, 611, 345, 466, 326, 1576, 11, 345, 47873, 284, 257, 1365, 2746, 780, 612, 338, 50767], "temperature": 0.0, "avg_logprob": -0.15742188579631303, "compression_ratio": 1.6900369003690037, "no_speech_prob": 0.0047852229326963425, "words": [{"word": " constantly", "start": 617.58, "end": 617.86, "probability": 0.9965511560440063}, {"word": " doing", "start": 617.86, "end": 618.22, "probability": 0.9973253011703491}, {"word": " that.", "start": 618.22, "end": 618.54, "probability": 0.9981667995452881}, {"word": " And", "start": 618.82, "end": 618.9, "probability": 0.9843797087669373}, {"word": " if", "start": 618.9, "end": 619.0, "probability": 0.9944820404052734}, {"word": " you", "start": 619.0, "end": 619.1, "probability": 0.9991812109947205}, {"word": " do", "start": 619.1, "end": 619.22, "probability": 0.9981354475021362}, {"word": " that", "start": 619.22, "end": 619.36, "probability": 0.9951167106628418}, {"word": " enough,", "start": 619.36, "end": 619.66, "probability": 0.9957749247550964}, {"word": " you", "start": 620.08, "end": 620.14, "probability": 0.9935512542724609}, {"word": " converge", "start": 620.14, "end": 620.74, "probability": 0.9099768400192261}, {"word": " to", "start": 620.74, "end": 621.0, "probability": 0.996359646320343}, {"word": " a", "start": 621.0, "end": 621.08, "probability": 0.9985860586166382}, {"word": " better", "start": 621.08, "end": 621.26, "probability": 0.9996664524078369}, {"word": " model", "start": 621.26, "end": 621.58, "probability": 0.996767520904541}, {"word": " because", "start": 621.58, "end": 622.16, "probability": 0.8104948401451111}, {"word": " there's", "start": 622.16, "end": 622.4, "probability": 0.9931629002094269}]}, {"id": 112, "seek": 61424, "start": 622.4, "end": 627.72, "text": " more data. Question in the back. Is there a something here that we're like essentially", "tokens": [50767, 517, 1366, 13, 18233, 287, 262, 736, 13, 1148, 612, 257, 1223, 994, 326, 356, 821, 588, 6986, 51047], "temperature": 0.0, "avg_logprob": -0.15742188579631303, "compression_ratio": 1.6900369003690037, "no_speech_prob": 0.0047852229326963425, "words": [{"word": " more", "start": 622.4, "end": 622.54, "probability": 0.9988890290260315}, {"word": " data.", "start": 622.54, "end": 622.84, "probability": 0.9975416660308838}, {"word": " Question", "start": 623.62, "end": 624.04, "probability": 0.35135167837142944}, {"word": " in", "start": 624.04, "end": 624.24, "probability": 0.9765223264694214}, {"word": " the", "start": 624.24, "end": 624.3, "probability": 0.9982492923736572}, {"word": " back.", "start": 624.3, "end": 624.74, "probability": 0.9941396713256836}, {"word": " Is", "start": 625.18, "end": 625.64, "probability": 0.08593878895044327}, {"word": " there", "start": 625.64, "end": 625.82, "probability": 0.9786491990089417}, {"word": " a", "start": 625.82, "end": 626.22, "probability": 0.597769021987915}, {"word": " something", "start": 626.22, "end": 626.44, "probability": 0.11990328878164291}, {"word": " here", "start": 626.44, "end": 626.7, "probability": 0.8955555558204651}, {"word": " that", "start": 626.7, "end": 626.9, "probability": 0.9463455677032471}, {"word": " we're", "start": 626.9, "end": 627.14, "probability": 0.9202385544776917}, {"word": " like", "start": 627.14, "end": 627.32, "probability": 0.8166650533676147}, {"word": " essentially", "start": 627.32, "end": 627.72, "probability": 0.42842772603034973}]}, {"id": 113, "seek": 61424, "start": 627.72, "end": 633.2, "text": " averaging the model weights across all of the individuals? And if so, is that appropriate?", "tokens": [51047, 20430, 262, 2746, 19590, 1973, 477, 286, 262, 3925, 30, 843, 611, 523, 11, 318, 326, 5035, 30, 51311], "temperature": 0.0, "avg_logprob": -0.15742188579631303, "compression_ratio": 1.6900369003690037, "no_speech_prob": 0.0047852229326963425, "words": [{"word": " averaging", "start": 627.72, "end": 628.48, "probability": 0.9582520127296448}, {"word": " the", "start": 628.48, "end": 628.96, "probability": 0.8182607293128967}, {"word": " model", "start": 628.96, "end": 629.32, "probability": 0.919602632522583}, {"word": " weights", "start": 629.32, "end": 629.64, "probability": 0.9689487814903259}, {"word": " across", "start": 629.64, "end": 630.08, "probability": 0.9917349815368652}, {"word": " all", "start": 630.08, "end": 630.5, "probability": 0.9745820760726929}, {"word": " of", "start": 630.5, "end": 630.8, "probability": 0.7086626291275024}, {"word": " the", "start": 630.8, "end": 630.94, "probability": 0.9626885652542114}, {"word": " individuals?", "start": 630.94, "end": 631.4, "probability": 0.6308577060699463}, {"word": " And", "start": 631.78, "end": 631.78, "probability": 0.6756659150123596}, {"word": " if", "start": 631.78, "end": 632.14, "probability": 0.830176591873169}, {"word": " so,", "start": 632.14, "end": 632.4, "probability": 0.9803398251533508}, {"word": " is", "start": 632.48, "end": 632.58, "probability": 0.9332752227783203}, {"word": " that", "start": 632.58, "end": 632.72, "probability": 0.9930278658866882}, {"word": " appropriate?", "start": 632.72, "end": 633.2, "probability": 0.33443233370780945}]}, {"id": 114, "seek": 61424, "start": 634.5800000000002, "end": 640.44, "text": " So the early works on federated learning do literally just average the weights. There are", "tokens": [51359, 1406, 262, 1903, 2499, 319, 28062, 515, 4673, 466, 7360, 655, 2811, 262, 19590, 13, 1318, 389, 51683], "temperature": 0.0, "avg_logprob": -0.15742188579631303, "compression_ratio": 1.6900369003690037, "no_speech_prob": 0.0047852229326963425, "words": [{"word": " So", "start": 634.5800000000002, "end": 635.1400000000001, "probability": 0.9247942566871643}, {"word": " the", "start": 635.1400000000001, "end": 635.7, "probability": 0.6801298260688782}, {"word": " early", "start": 635.7, "end": 636.02, "probability": 0.9862903952598572}, {"word": " works", "start": 636.02, "end": 636.32, "probability": 0.9833499193191528}, {"word": " on", "start": 636.32, "end": 636.48, "probability": 0.9808804392814636}, {"word": " federated", "start": 636.48, "end": 636.88, "probability": 0.9116668701171875}, {"word": " learning", "start": 636.88, "end": 637.16, "probability": 0.995322048664093}, {"word": " do", "start": 637.16, "end": 637.44, "probability": 0.7825886011123657}, {"word": " literally", "start": 637.44, "end": 637.92, "probability": 0.9821622967720032}, {"word": " just", "start": 637.92, "end": 638.54, "probability": 0.9919567108154297}, {"word": " average", "start": 638.54, "end": 639.1, "probability": 0.9824104309082031}, {"word": " the", "start": 639.1, "end": 639.38, "probability": 0.9963786005973816}, {"word": " weights.", "start": 639.38, "end": 639.66, "probability": 0.9911638498306274}, {"word": " There", "start": 640.06, "end": 640.1, "probability": 0.9934281706809998}, {"word": " are", "start": 640.1, "end": 640.44, "probability": 0.9986463189125061}]}, {"id": 115, "seek": 64044, "start": 640.44, "end": 647.94, "text": " more advanced schemes, especially ones that can help with data privacy. And so sometimes it's a", "tokens": [50371, 517, 6190, 16546, 11, 2592, 3392, 326, 460, 1037, 351, 1366, 6782, 13, 843, 523, 3360, 340, 338, 257, 50731], "temperature": 0.0, "avg_logprob": -0.13082560300827026, "compression_ratio": 1.5658536585365854, "no_speech_prob": 0.0028317864052951336, "words": [{"word": " more", "start": 640.44, "end": 640.86, "probability": 0.46772417426109314}, {"word": " advanced", "start": 640.86, "end": 642.14, "probability": 0.9837026596069336}, {"word": " schemes,", "start": 642.14, "end": 642.82, "probability": 0.9962007403373718}, {"word": " especially", "start": 643.18, "end": 643.44, "probability": 0.9978047013282776}, {"word": " ones", "start": 643.44, "end": 643.76, "probability": 0.9930103421211243}, {"word": " that", "start": 643.76, "end": 643.92, "probability": 0.9986327290534973}, {"word": " can", "start": 643.92, "end": 644.04, "probability": 0.9807510375976562}, {"word": " help", "start": 644.04, "end": 644.32, "probability": 0.9994526505470276}, {"word": " with", "start": 644.32, "end": 644.72, "probability": 0.998551070690155}, {"word": " data", "start": 644.72, "end": 645.12, "probability": 0.9948779344558716}, {"word": " privacy.", "start": 645.12, "end": 645.74, "probability": 0.9991115927696228}, {"word": " And", "start": 646.96, "end": 646.96, "probability": 0.8007274270057678}, {"word": " so", "start": 646.96, "end": 647.24, "probability": 0.9884938597679138}, {"word": " sometimes", "start": 647.24, "end": 647.6, "probability": 0.8619112372398376}, {"word": " it's", "start": 647.6, "end": 647.88, "probability": 0.9260973036289215}, {"word": " a", "start": 647.88, "end": 647.94, "probability": 0.9986134767532349}]}, {"id": 116, "seek": 64044, "start": 647.94, "end": 651.92, "text": " bit more complicated, but at a high level, you should think of it that way is that you're just", "tokens": [50731, 1643, 517, 8253, 11, 475, 379, 257, 1029, 1241, 11, 345, 815, 892, 286, 340, 326, 835, 318, 326, 345, 821, 655, 50939], "temperature": 0.0, "avg_logprob": -0.13082560300827026, "compression_ratio": 1.5658536585365854, "no_speech_prob": 0.0028317864052951336, "words": [{"word": " bit", "start": 647.94, "end": 648.06, "probability": 0.9988032579421997}, {"word": " more", "start": 648.06, "end": 648.24, "probability": 0.9994231462478638}, {"word": " complicated,", "start": 648.24, "end": 648.8, "probability": 0.9995360374450684}, {"word": " but", "start": 649.16, "end": 649.34, "probability": 0.9979058504104614}, {"word": " at", "start": 649.34, "end": 649.54, "probability": 0.9960852861404419}, {"word": " a", "start": 649.54, "end": 649.64, "probability": 0.9986018538475037}, {"word": " high", "start": 649.64, "end": 649.84, "probability": 0.9978896975517273}, {"word": " level,", "start": 649.84, "end": 650.06, "probability": 0.9913163185119629}, {"word": " you", "start": 650.2, "end": 650.3, "probability": 0.9989323019981384}, {"word": " should", "start": 650.3, "end": 650.48, "probability": 0.9958114624023438}, {"word": " think", "start": 650.48, "end": 650.66, "probability": 0.9993096590042114}, {"word": " of", "start": 650.66, "end": 650.82, "probability": 0.9946274161338806}, {"word": " it", "start": 650.82, "end": 650.92, "probability": 0.9990737438201904}, {"word": " that", "start": 650.92, "end": 651.06, "probability": 0.9910362958908081}, {"word": " way", "start": 651.06, "end": 651.26, "probability": 0.9995323419570923}, {"word": " is", "start": 651.26, "end": 651.5, "probability": 0.3369229733943939}, {"word": " that", "start": 651.5, "end": 651.62, "probability": 0.998108983039856}, {"word": " you're", "start": 651.62, "end": 651.8, "probability": 0.9962362349033356}, {"word": " just", "start": 651.8, "end": 651.92, "probability": 0.9983805418014526}]}, {"id": 117, "seek": 64044, "start": 651.92, "end": 656.78, "text": " averaging the parameters. There's some wrinkles in there, but that's how you should think of it.", "tokens": [50939, 20430, 262, 10007, 13, 1318, 338, 617, 49391, 287, 612, 11, 475, 326, 338, 703, 345, 815, 892, 286, 340, 13, 51175], "temperature": 0.0, "avg_logprob": -0.13082560300827026, "compression_ratio": 1.5658536585365854, "no_speech_prob": 0.0028317864052951336, "words": [{"word": " averaging", "start": 651.92, "end": 652.4, "probability": 0.9960350394248962}, {"word": " the", "start": 652.4, "end": 652.78, "probability": 0.9881254434585571}, {"word": " parameters.", "start": 652.78, "end": 653.3, "probability": 0.993274450302124}, {"word": " There's", "start": 653.56, "end": 653.9, "probability": 0.9005637764930725}, {"word": " some", "start": 653.9, "end": 654.06, "probability": 0.9650217890739441}, {"word": " wrinkles", "start": 654.06, "end": 654.82, "probability": 0.7770819067955017}, {"word": " in", "start": 654.82, "end": 655.16, "probability": 0.9975519776344299}, {"word": " there,", "start": 655.16, "end": 655.34, "probability": 0.999289870262146}, {"word": " but", "start": 655.36, "end": 655.54, "probability": 0.9990720748901367}, {"word": " that's", "start": 655.54, "end": 655.86, "probability": 0.9992988109588623}, {"word": " how", "start": 655.86, "end": 655.98, "probability": 0.9995614886283875}, {"word": " you", "start": 655.98, "end": 656.12, "probability": 0.9996557235717773}, {"word": " should", "start": 656.12, "end": 656.26, "probability": 0.9993740916252136}, {"word": " think", "start": 656.26, "end": 656.44, "probability": 0.9991462230682373}, {"word": " of", "start": 656.44, "end": 656.58, "probability": 0.9991291165351868}, {"word": " it.", "start": 656.58, "end": 656.78, "probability": 0.9983493089675903}]}, {"id": 118, "seek": 64044, "start": 658.0600000000001, "end": 660.1, "text": " Great. Thank you very much, Zach.", "tokens": [51247, 3878, 13, 6952, 345, 845, 881, 11, 18825, 13, 51347], "temperature": 0.0, "avg_logprob": -0.13082560300827026, "compression_ratio": 1.5658536585365854, "no_speech_prob": 0.0028317864052951336, "words": [{"word": " Great.", "start": 658.0600000000001, "end": 658.46, "probability": 0.4408349096775055}, {"word": " Thank", "start": 659.32, "end": 659.4, "probability": 0.9937084913253784}, {"word": " you", "start": 659.4, "end": 659.5, "probability": 0.9991137385368347}, {"word": " very", "start": 659.5, "end": 659.66, "probability": 0.9912623167037964}, {"word": " much,", "start": 659.66, "end": 659.9, "probability": 0.9990845918655396}, {"word": " Zach.", "start": 660.06, "end": 660.1, "probability": 0.40926456451416016}]}], "language": "en"}}