Split (1)

train · 370 rows

id stringlengths 11 11	channel stringclasses 2 values	channel_id stringclasses 2 values	title stringlengths 12 100	categories list	tags list	description stringlengths 66 5k	text stringlengths 577 90.4k	segments list
z4lAlVRwbrc	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Author Interview - Improving Intrinsic Exploration with Language Abstractions	[ "Science & Technology" ]	[ "" ]	#reinforcementlearning #ai #explained This is an interview with Jesse Mu, first author of the paper. Original Paper Review: https://youtu.be/NeGJAUSQEJI Exploration is one of the oldest challenges for Reinforcement Learning algorithms, with no clear solution to date. Especially in environments with sparse rewards, ag...	Hello, this is an interview with Jesse Mu, who is the first author of the paper improving intrinsic exploration with language abstractions. This paper is really cool because it combines the knowledge that is inherent in language with the problem of exploration in reinforcement learning. I've made a comprehensive revie...	[ { "start": 0, "end": 10.56, "text": " Hello, this is an interview with Jesse Mu, who is the first author of the paper improving" }, { "start": 10.56, "end": 13.84, "text": " intrinsic exploration with language abstractions." }, { "start": 13.84, "end": 18.44, "text": " Th...
ZTs_mXwMCs8	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Galactica: A Large Language Model for Science (Drama & Paper Review)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "galactica", "meta", "meta ai", "facebook ai", "ai science", "galactica ai", "galactica model", "yann lecun", "research", "fair", "deep learning tutorial", "w...	#ai #galactica #meta Galactica is a language model trained on a curated corpus of scientific documents, such as papers, knowledge bases, reviews, and other articles. The model can be used in a generative fasion to assist scientific writing, do reference prediction, and much more, including a new approach to do step-by...	Hello, this video starts out with a review of the drama around the public demo of the Galactica model and then goes into a paper review. If you're not in the mood for any drama, skip ahead about 16 minutes and you'll be fine. Hello there. Galactica is a model, a language model by MetaAI that is trained specifically on...	[ { "start": 0, "end": 5.24, "text": " Hello, this video starts out with a review of the drama around the public demo of the" }, { "start": 5.24, "end": 8.92, "text": " Galactica model and then goes into a paper review." }, { "start": 8.92, "end": 14.72, "text": " If you're...
n1SXlK5rhR8	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[Drama] Yann LeCun against Twitter on Dataset Bias	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "ylc", "yann", "lecun", "convnet", "face", "pulse", "github", "colab", "jeff dean", "hardmaru", "charles sutton", "soumith", "meredith", "timnit", "bias...	Yann LeCun points out an instance of dataset bias and proposes a sensible solution. People are not happy about it. Original Tweet: https://twitter.com/ylecun/status/1274782757907030016 ERRATA: - My specific example of the L1 regularizer wrt to Porsches and Ferraris does not actually work in this particular case. What...	Hi there! So you may have seen this already. There's a CVPR paper called Pulse. And what it does is it's a method to up sample a pixelated image in a way that makes it look realistic, but also that the again down sampled variant matches the original down sampled image. So it's kind of a cycle consistency loss together...	[ { "start": 0, "end": 6.32, "text": " Hi there! So you may have seen this already. There's a CVPR paper called Pulse. And what it" }, { "start": 6.32, "end": 12.88, "text": " does is it's a method to up sample a pixelated image in a way that makes it look realistic," }, { "start":...
BK3rv0MQMwY	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[News] The Siraj Raval Controversy	[ "Science & Technology" ]	[ "machine learning", "siraj", "controversy", "scam", "scammer", "fraud", "plagiarism", "plagiarized", "course", "refund", "policy", "ai", "online", "hype", "credit", "attribution", "paper", "scandal", "news", "twitter", "neural qubit", "intellectual property" ]	Popular ML YouTuber Siraj Raval is in the middle of not just one, but two controversies: First, a lot of students of his 200$ online-course have accused him of breaking major promises he made when advertising the course and denying them refunds. Second, his paper on "The Neural Qubit" appears to be plagiarized almost v...	There is a massive controversy going on right now and in the middle is Siraj Raval, a prominent YouTuber. So today I'll just be actually shortly reporting on this, not giving too much opinion, just kind of stating what's up in a very high level overview. Because if you haven't heard of this, I think it's important tha...	[ { "start": 0, "end": 7, "text": " There is a massive controversy going on right now and in the middle is Siraj Raval, a prominent" }, { "start": 7.6000000000000005, "end": 14.6, "text": " YouTuber. So today I'll just be actually shortly reporting on this, not giving too much opinion," ...
U0mxx7AoNz0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Player of Games: All the games, one algorithm! (w/ author Martin Schmid)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "reinforcement learning", "ai for go", "ai go", "ai chess", "chess ai", "stockfish", "alphazero", "alpha zero", "muzero", "player of games", "pog", "deepmind"...	#playerofgames #deepmind #alphazero Special Guest: First author Martin Schmid (https://twitter.com/Lifrordi) Games have been used throughout research as testbeds for AI algorithms, such as reinforcement learning agents. However, different types of games usually require different solution approaches, such as AlphaZero ...	Hello everyone, today is a special day. I'm here, as you can see, not alone, not by myself as usual. I'm joined by Martin Schmidt, who is the first author of the paper called Player of Games. This is joint work with others by DeepMind and I have to say it's a very in-depth paper. It presents an algorithm called Player...	[ { "start": 0, "end": 7.28, "text": " Hello everyone, today is a special day. I'm here, as you can see, not alone, not by myself as usual." }, { "start": 7.28, "end": 13.84, "text": " I'm joined by Martin Schmidt, who is the first author of the paper called Player of Games." }, { ...
fvctpYph8Pc	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Do ImageNet Classifiers Generalize to ImageNet? (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "imagenet", "cifar10", "cifar10.1", "generalization", "overfitting", "mturk", "arxiv", "vision", "models", "research", "hardness", "accuracy", "classifier", "resnet" ]	Has the world overfitted to ImageNet? What if we collect another dataset in exactly the same fashion? This paper gives a surprising answer! Paper: https://arxiv.org/abs/1902.10811 Data: https://github.com/modestyachts/ImageNetV2 Abstract: We build new test sets for the CIFAR-10 and ImageNet datasets. Both benchmarks ...	Hi there today we're looking at to do image net classifiers Generalized to image net by Benjamin wrecked Rebecca are Olaf's Ludwig Schmidt and Vyshal Shankar So the premise of this paper is pretty simple We've been training models on image net now for a while Almost ten years to be exact image net is this data set wit...	[ { "start": 0, "end": 3.3000000000000003, "text": " Hi there today we're looking at to do image net classifiers" }, { "start": 3.6, "end": 9.78, "text": " Generalized to image net by Benjamin wrecked Rebecca are Olaf's Ludwig Schmidt and Vyshal Shankar" }, { "start": 10.1, "en...
PZypP7PiKi0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Gradient Surgery for Multi-Task Learning	[ "Science & Technology" ]	[ "deep learning", "machine learning", "neural networks", "multi task", "conflicting gradients", "magnitudes", "adam", "sgd", "momentum", "optimization", "projection" ]	Multi-Task Learning can be very challenging when gradients of different tasks are of severely different magnitudes or point into conflicting directions. PCGrad eliminates this problem by projecting conflicting gradients while still retaining optimality guarantees. https://arxiv.org/abs/2001.06782 Abstract: While deep...	Hi there, today we're looking at gradient surgery for multitask learning by Tianhe Yu, Saurabh Kumar, Abhishek Gupta, Sergei Levine, Carole Hausmann and Chelsea Finn. So in this paper, the concern is a thing called multitask learning. Now what is multitask learning? So this has some very subtle distinctions from other...	[ { "start": 0, "end": 7, "text": " Hi there, today we're looking at gradient surgery for multitask learning by Tianhe Yu," }, { "start": 7, "end": 15.08, "text": " Saurabh Kumar, Abhishek Gupta, Sergei Levine, Carole Hausmann and Chelsea Finn." }, { "start": 15.08, "end": 22.2...
Z3knUzwuIgo	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	One Model For All The Tasks - BLIP (Author Interview)	[ "Science & Technology" ]	[]	#blip #interview #salesforce Paper Review Video: https://youtu.be/X2k7n4FuI7c Sponsor: Assembly AI https://www.assemblyai.com/?utm_source=youtube&utm_medium=social&utm_campaign=yannic2 This is an interview with Junnan Li and Dongxu Li, authors of BLIP and members of Salesforce research. Cross-modal pre-training has b...	Hello, this is an interview with the authors of the blip paper. If you haven't seen it, I've made a review video of the paper itself. Be sure to check that out. The authors have seen that and are directly able to respond to it. So we all start on an even footing. It's very cool to have the authors on and this intervie...	[ { "start": 0, "end": 9.200000000000001, "text": " Hello, this is an interview with the authors of the blip paper." }, { "start": 9.200000000000001, "end": 13.64, "text": " If you haven't seen it, I've made a review video of the paper itself." }, { "start": 13.64, "end": 14.8,...
3Tqp_B2G6u0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Blockwise Parallel Decoding for Deep Autoregressive Models	[ "Science & Technology" ]	[ "machine learning", "deep learning", "transformers", "nlp", "natural language processing", "ai", "artificial intelligence", "google brain", "autoregressive", "greedy decoding", "inference", "language model", "speedup" ]	https://arxiv.org/abs/1811.03115 Abstract: Deep autoregressive sequence-to-sequence models have demonstrated impressive performance across a wide variety of tasks in recent years. While common architecture classes such as recurrent, convolutional, and self-attention networks make different trade-offs between the amoun...	Hi there, today we'll look at blockwise parallel decoding for deep autoregressive models by Mitchell Stern, Noam Shazir and Jakob Uschkordei of UC Berkeley and Google Brain. So this is a bit more of an engineering paper than usual, which I find cool. It's basically an engineering trick to get these autoregressive mode...	[ { "start": 0, "end": 6.640000000000001, "text": " Hi there, today we'll look at blockwise parallel decoding for deep autoregressive models by" }, { "start": 6.640000000000001, "end": 15.200000000000001, "text": " Mitchell Stern, Noam Shazir and Jakob Uschkordei of UC Berkeley and Google ...
8wkgDnNxiVs	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	POET: Endlessly Generating Increasingly Complex and Diverse Learning Environments and Solutions	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "evolution", "reinforcement learning", "neat", "open-ended", "never ending", "population", "bipedal walker" ]	From the makers of Go-Explore, POET is a mixture of ideas from novelty search, evolutionary methods, open-ended learning and curriculum learning. https://arxiv.org/abs/1901.01753 Abstract: While the history of machine learning so far largely encompasses a series of problems posed by researchers and algorithms that le...	Alright, so what you're seeing here are solutions found to this bipedal walker problem by a new algorithm called PoET. So as you might guess, the challenge is to keep this little thing here walking to the right as far as you can while it encounters various obstacles. And it is and remains a challenging reinforcement l...	[ { "start": 0, "end": 6.88, "text": " Alright, so what you're seeing here are solutions found to this bipedal walker problem by a" }, { "start": 6.88, "end": 10.52, "text": " new algorithm called PoET." }, { "start": 10.52, "end": 16.84, "text": " So as you might guess, th...
We20YSAJZSE	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	MuZero: Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model	[ "Science & Technology" ]	[ "ml", "ai", "machine learning", "reinforcement learning", "deep rl", "deepmind", "google", "alphago", "alphazero", "value function", "policy", "artificial intelligence", "rl", "deep reinforcement learning", "model-free", "model-based", "environment model", "hidden representation", ...	MuZero harnesses the power of AlphaZero, but without relying on an accurate environment model. This opens up planning-based reinforcement learning to entirely new domains, where such environment models aren't available. The difference to previous work is that, instead of learning a model predicting future observations,...	Hi there! Today we're looking at mastering Atari Go, Chess and Shogi by planning with a learned model by Julian Schrittweiser and people generally from DeepMind. So this paper is an extension to AlphaZero, the kind of famous algorithm that learned to play Go and Chess simply by playing itself and the kind of cool thin...	[ { "start": 0, "end": 5.82, "text": " Hi there! Today we're looking at mastering Atari Go, Chess and Shogi by" }, { "start": 5.82, "end": 12.120000000000001, "text": " planning with a learned model by Julian Schrittweiser and people generally from" }, { "start": 12.120000000000001...
i-J4T3uLC9M	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	VOS: Learning What You Don't Know by Virtual Outlier Synthesis (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "paper explained", "virtual outliers", "how to detect outliers", "deep learning outliers", "deep learning outlier detection", "vos", "deep learning energy", "latent s...	#vos #outliers #deeplearning Sponsor: Assembly AI Check them out here: https://www.assemblyai.com/?utm_source=youtube&utm_medium=social&utm_campaign=yannic1 Outliers are data points that are highly unlikely to be seen in the training distribution, and therefore deep neural networks have troubles when dealing with them...	Outliers, we all know them, we all hate them. How can these data points just be out of distribution, not in the training data, things that we haven't seen before, things that we don't even expect? Well, they suck. So today we're going to look at what you can do about it. Specifically, we're going to look at the paper ...	[ { "start": 0, "end": 12.8, "text": " Outliers, we all know them, we all hate them. How can these data points just be out of distribution," }, { "start": 12.8, "end": 19.2, "text": " not in the training data, things that we haven't seen before, things that we don't even expect?" }, { ...
lqtlua-Ylts	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	State-of-Art-Reviewing: A Radical Proposal to Improve Scientific Publication	[ "Science & Technology" ]	[ "deep learning", "machine learning", "nlp", "natural language processing", "arxiv", "attention", "peer review", "automate", "distributed", "scalable", "neurips", "score", "objective" ]	Peer Review is outdated and ineffective. SOAR is a new and revolutionary way to distribute scientific reviewing and scale to the new age of faster, better and more significant research. https://arxiv.org/abs/2003.14415 Abstract: Peer review forms the backbone of modern scientific manuscript evaluation. But after two ...	Hi everyone. Today we're looking at state-of-the-art reviewing a radical proposal to improve scientific publication. This has been on my mind for a while. The review process for modern science, especially machine learning, is just broken. I've spoken numerous times about the fact that we need to replace it with a bett...	[ { "start": 0, "end": 8.6, "text": " Hi everyone. Today we're looking at state-of-the-art reviewing a radical proposal to improve scientific publication." }, { "start": 8.6, "end": 17.8, "text": " This has been on my mind for a while. The review process for modern science, especially mach...
xJrKIPwVwGM	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Rethinking Attention with Performers (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "nlp", "natural language processing", "natural language understanding", "data science", "transformer", "attention", "attention mechanism", "transformers", "attentio...	#ai #research #attention Transformers have huge memory and compute requirements because they construct an Attention matrix, which grows quadratically in the size of the input. The Performer is a model that uses random positive orthogonal features to construct an unbiased estimator to the Attention matrix and obtains a...	Hi there, today we'll look at rethinking attention with performers by researchers of Google, the University of Cambridge, DeepMind and the Alan Turing Institute. This paper is yet another paper in the quest to make transformers more performant and what better name to give to a technique than the performer. So the perf...	[ { "start": 0, "end": 7.640000000000001, "text": " Hi there, today we'll look at rethinking attention with performers by researchers of Google," }, { "start": 7.640000000000001, "end": 12.280000000000001, "text": " the University of Cambridge, DeepMind and the Alan Turing Institute." },...
ccBMRryxGog	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Sparse Expert Models (Switch Transformers, GLAM, and more... w/ the Authors)	[ "Science & Technology" ]	[]	#nlp #sparsity #transformers This video is an interview with Barret Zoph and William Fedus of Google Brain about Sparse Expert Models. Sparse Expert models have been hugely successful at distributing parts of models, mostly Transformers, across large array of machines and use a routing function to effectively route si...	Hello, today I'm having an interview about the topic of sparse experts. Now, ironically, the people are absolute experts in this type of models. These models, they are huge, they're usually language models, but they don't have to be they're usually transformers, but they don't have to be what they do have in common is...	[ { "start": 0, "end": 5.2, "text": " Hello, today I'm having an interview about the topic of sparse experts. Now, ironically," }, { "start": 5.2, "end": 11.120000000000001, "text": " the people are absolute experts in this type of models. These models, they are huge, they're" }, { ...
3jT1qJ8ETzk	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	SupSup: Supermasks in Superposition (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "supsup", "supermasks", "lottery ticket", "lottery ticket hypothesis", "gradient", "entropy", "surplus", "superfluous neurons", "lifelong learning", "multitask le...	Supermasks are binary masks of a randomly initialized neural network that result in the masked network performing well on a particular task. This paper considers the problem of (sequential) Lifelong Learning and trains one Supermask per Task, while keeping the randomly initialized base network constant. By minimizing t...	Hi there, today we'll look at super masks in superposition by Mitchell Wirtzman, Vivek Ramanujan at AL. So on a high level this paper tackles the problem of sequentially learning many many tasks without catastrophic forgetting by leveraging these things called super masks. A super mask is basically a binary mask that ...	[ { "start": 0, "end": 7, "text": " Hi there, today we'll look at super masks in superposition by Mitchell Wirtzman, Vivek Ramanujan at AL." }, { "start": 7, "end": 15, "text": " So on a high level this paper tackles the problem of sequentially learning many many tasks without catastrophic...
kEhEbVZQwjM	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[YTalks] Siraj Raval - Stories about YouTube, Plagiarism, and the Dangers of Fame (Interview)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "siraj", "siraj raval", "ml youtube", "fame", "youtuber life", "what happened to siraj", "siraj raval plagiarism", "siraj raval interview", "siraj raval coursera", ...	#ytalks #siraj #plagiarism A conversation with Siraj Raval about his journey on YouTube, and the perils of fame. OUTLINE: 0:00 - Intro 1:30 - Welcome 3:15 - Starting out: From Economics to YouTube 13:00 - More Views: Plagiarizing Video Content 23:30 - One Step Up: Copying A Research Paper 29:15 - Was there another wa...	The following is a conversation with Siraj Ruval. Siraj has one of the largest channels in the machine learning YouTube space. Over 700,000 people are subscribed to him as of this date. Siraj pumped out lots and lots of videos on topics such as coding tutorials, explaining beginners concept in machine learning and in ...	[ { "start": 0, "end": 6.2, "text": " The following is a conversation with Siraj Ruval. Siraj has one of the largest" }, { "start": 6.2, "end": 11.16, "text": " channels in the machine learning YouTube space. Over 700,000 people are" }, { "start": 11.16, "end": 18.52, "text...
qeEO2GECQk0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Evaluating NLP Models via Contrast Sets	[ "Science & Technology" ]	[ "deep learning", "machine learning", "nlp", "natural language processing", "arxiv", "attention", "evaluation", "cheat", "easy", "hard", "adversarial", "counterfactual", "hand-crafted", "test set", "supervised" ]	Current NLP models are often "cheating" on supervised learning tasks by exploiting correlations that arise from the particularities of the dataset. Therefore they often fail to learn the original intent of the dataset creators. This paper argues that NLP models should be evaluated on Contrast Sets, which are hand-craft...	Hi there! Today we're looking at evaluating NLP models via contrast sets. These are too many authors from too many places for me to read out. We'll just jump right into the problem. What is the problem? Let's jump into the solution. Here you see a visual question answering task. Visual question answering in this case....	[ { "start": 0, "end": 5.68, "text": " Hi there! Today we're looking at evaluating NLP models via contrast sets." }, { "start": 5.68, "end": 12.8, "text": " These are too many authors from too many places for me to read out." }, { "start": 12.8, "end": 22.32, "text": " We'l...
tjbEVY5XIk0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Divide-and-Conquer Monte Carlo Tree Search For Goal-Directed Planning (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "rl", "reinforcement learning", "deep rl", "planning", "alphago", "alphazero", "alpha go", "alpha zero", "mcts", "monte carlo", "tree search", "subdivision", ...	When AI makes a plan it usually does so step by step, forward in time. But often it is beneficial to define intermediate goals to divide a large problem into easier sub-problems. This paper proposes a generalization of MCTS that searches not for the best next actions to take, but for the best way to sub-divide the prob...	Hi there! What you're seeing here is a Divide and Conquer Monte Carlo Tree Search in action. This is a planning algorithm that plans in a kind of an unconventional fashion. So we're going to explore this today in this paper. Divide and Conquer Monte Carlo Tree Search for Goal-Directed Planning by Gian Battista Parasco...	[ { "start": 0, "end": 5.64, "text": " Hi there! What you're seeing here is a Divide and Conquer Monte Carlo Tree" }, { "start": 5.64, "end": 12.64, "text": " Search in action. This is a planning algorithm that plans in a kind of an" }, { "start": 12.64, "end": 17.52, "text...
SPOqoI0zOPQ	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[ML News] AI-generated patent approved \| Germany gets an analog to OpenAI \| ML cheats video games	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "deep learning tutorial", "what is deep learning", "introduction to deep learning", "ai inventor", "dabus", "thaler", "steve thaler", "stephen thaler", "ai patent",...	#mlnews #dabus #alephalpha OUTLINE: 0:00 - Intro 0:20 - Sponsor: Weights & Biases 3:45 - AI legally recognized as patent inventor 8:35 - Alpeh Alpha raises USD 27Mio to build European OpenAI 10:20 - AMP advances AI aided recycling 11:20 - DeepMind builds XLand RL environment 13:15 - Cognitive Behavioral Therapy as an ...	An AI is now officially listed as the inventor in a patent. Aleph Alpha raises $27 million to build Europe's open AI and an open source replication of Dalí is released. Welcome to ML News. All right, before we get into all the stuff, this video is sponsored by weight and biases. weights and biases is a one stop shop f...	[ { "start": 0, "end": 7.2, "text": " An AI is now officially listed as the inventor in a patent. Aleph Alpha raises $27 million to" }, { "start": 7.2, "end": 13.84, "text": " build Europe's open AI and an open source replication of Dalí is released. Welcome to ML News." }, { "star...
Lg97gWXsiQ4	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Resolution-robust Large Mask Inpainting with Fourier Convolutions (w/ Author Interview)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "lama", "inpainting", "gan", "adversarial", "loss function", "fourier transform", "fft", "fast fourier transform", "fourier convolution", "fast fourier convolutio...	#lama #inpainting #deeplearning At the end of the video is an interview with the paper authors! LaMa is a system that is amazing at removing foreground objects from images, especially when those objects cover a large part of the image itself. LaMa is specifically trained to reconstruct large masked areas and includes ...	Hello there, today we're looking at resolution robust large mask in painting with Fourier convolutions also called LAMA by the Samsung AI Center, Samsung Research, EPFL and the Skolkovo Institute of Science and Technology. This is a special paper review because I'm only going to introduce the paper briefly, maybe 15-2...	[ { "start": 0, "end": 5.36, "text": " Hello there, today we're looking at resolution robust large mask in painting with Fourier" }, { "start": 5.36, "end": 12.96, "text": " convolutions also called LAMA by the Samsung AI Center, Samsung Research, EPFL and the Skolkovo" }, { "start...
W3mrgqtm5R4	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[ML News] BLOOM: 176B Open-Source \| Chinese Brain-Scale Computer \| Meta AI: No Language Left Behind	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "bloom", "nlp", "gpt3", "gpt 3", "gpt-3", "eleuther ai", "eleutherai", "bigscience", "bigsciencew", "big science", "huggingface", "hugging face", "yalm", ...	#mlnews #bloom #ai Today we look at all the recent giant language models in the AI world! OUTLINE: 0:00 - Intro 0:55 - BLOOM: Open-Source 176B Language Model 5:25 - YALM 100B 5:40 - Chinese Brain-Scale Supercomputer 7:25 - Meta AI Translates over 200 Languages 10:05 - Reproducibility Crisis Workshop 10:55 - AI21 Rai...	Bloom finishes training and is now released as the biggest open source language model to date. A new Chinese supercomputer is allegedly able to compute brain scale AI models. And both Ian Goodfellow and Andrej Karpati leave their jobs. Welcome to ML News. Hello and welcome everyone to ML News rather ML old I've been g...	[ { "start": 0.48, "end": 6.32, "text": " Bloom finishes training and is now released as the biggest open source language model to date." }, { "start": 6.88, "end": 12.88, "text": " A new Chinese supercomputer is allegedly able to compute brain scale AI models." }, { "start": 13.52...
RrBapqCPnmE	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[ML Coding Tips] Separate Computation & Plotting using locals	[ "Science & Technology" ]	[ "deep learning", "machine learning", "coding", "research", "engineering", "ipython", "colab", "notebook", "locals" ]	Here's a lazy way to separate computation and subsequent analysis in a notebook without the overhead of manually saving local variables. WARNING: Don't do this in a serious project. Links: YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher BitChute: https://www.bitchute.com/channel...	Hi there! So today I just wanted to bring you a quick coding tip that I often encounter in my daily machine learning research or life that might not be super common in let's say traditional software engineering or elsewhere. So often I have a bunch of, let's say I have a bunch of models right, and I use these IPython ...	[ { "start": 0, "end": 5.5600000000000005, "text": " Hi there! So today I just wanted to bring you a quick coding tip that I often" }, { "start": 5.5600000000000005, "end": 11.76, "text": " encounter in my daily machine learning research or life that might not be super" }, { "start...
vfBAUYpMCTU	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Unsupervised Brain Models - How does Deep Learning inform Neuroscience? (w/ Patrick Mineault)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "xcorr", "patrick mineault", "unsupervised models", "neuroscience", "neuroscience and deep learning", "deep learning brain", "machine learning brain", "brain models",...	#deeplearning #brain #neuroscience Originally, Deep Learning sprang into existence inspired by how the brain processes information, but the two fields have diverged ever since. However, given that deep models can solve many perception tasks with remarkable accuracy, is it possible that we might be able to learn someth...	Hello there! Today I'm interviewing Patrick Minot, who has a PhD from McGill and did a postdoc at UCLA. He's an independent scientist and a neural data scientist. His interests are neuroscience and the connection to machine learning. He has an awesome blog called XCore, which I guess is pronounced cross correlation, b...	[ { "start": 0, "end": 8.48, "text": " Hello there! Today I'm interviewing Patrick Minot, who has a PhD from McGill and did a postdoc at UCLA." }, { "start": 8.48, "end": 14.88, "text": " He's an independent scientist and a neural data scientist. His interests are neuroscience and" }, ...
i4H0kjxrias	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Reformer: The Efficient Transformer	[ "Science & Technology" ]	[ "deep learning", "machine learning", "nlp", "natural language processing", "machine translation", "arxiv", "google", "attention mechanism", "attention", "transformer", "seq2seq", "bert", "memory", "lsh", "locality sensitive hashing", "reversible", "revertible", "flow", "long sequ...	The Transformer for the masses! Reformer solves the biggest problem with the famous Transformer model: Its huge resource requirements. By cleverly combining Locality Sensitive Hashing and ideas from Reversible Networks, the classically huge footprint of the Transformer is drastically reduced. Not only does that mean th...	Hi there! Today we'll look at Reformer, the efficient transformer by Nikita Kitaev, Lukas Kaiser and Anselm Levskaia. This is a paper that tries to reduce the extreme resource requirements of the transformer model. Now if you haven't seen the transformer model before, that's this thing, I suggest you go watch for exam...	[ { "start": 0, "end": 5.84, "text": " Hi there! Today we'll look at Reformer, the efficient transformer by Nikita" }, { "start": 5.84, "end": 13.72, "text": " Kitaev, Lukas Kaiser and Anselm Levskaia. This is a paper that tries to reduce the" }, { "start": 13.72, "end": 18.6, ...
yFAuXmcGk2Y	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	SingularityNET - A Decentralized, Open Market and Network for AIs (Whitepaper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "singularity", "singularitynet", "agi", "ben goertzel", "goertzel", "hanson", "hanson robotics", "sophia", "network", "api", "offercoin", "offernetworks", "...	#ai #research #blockchain Big Tech is currently dominating the pursuit of ever more capable AI. This happens behind closed doors and results in a monopoly of power. SingularityNET is an open, decentralized network where anyone can offer and consume AI services, and where AI agents can interlink with each other to prov...	Hi there. Today we'll look at SingularityNet, the global AI marketplace, as it is advertised on their website. Specifically, we're going to look at the SingularityNet white paper 2.0, as it appeared in 2019. So it's version two, version one, I think appeared in 2017. So SingularityNet is a, as it says, a global AI mar...	[ { "start": 0, "end": 7.140000000000001, "text": " Hi there. Today we'll look at SingularityNet, the global AI marketplace, as it is advertised" }, { "start": 7.140000000000001, "end": 12.540000000000001, "text": " on their website. Specifically, we're going to look at the SingularityNet ...
LMb5tvW-UoQ	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Discovering Symbolic Models from Deep Learning with Inductive Biases (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "graph networks", "graph neural networks", "gnn", "physics", "newtonian", "hamiltonian", "dynamics", "cosmology", "dark matter", "symbolic regression", "edge", ...	Neural networks are very good at predicting systems' numerical outputs, but not very good at deriving the discrete symbolic equations that govern many physical systems. This paper combines Graph Networks with symbolic regression and shows that the strong inductive biases of these models can be used to derive accurate s...	Hi there, today we're looking at discovering symbolic models from deep learning with inductive biases by Miles Cranmer, Alvaro Sanchez-Gonzalez, Peter Pitalia, Rui Xu, Kyle Cranmer, David Spurgill and Shirley Ho. So this paper on a high level, it uses graph neural networks to fit a dataset of observations of a physica...	[ { "start": 0, "end": 5.28, "text": " Hi there, today we're looking at discovering symbolic models from deep learning with inductive" }, { "start": 5.28, "end": 11.98, "text": " biases by Miles Cranmer, Alvaro Sanchez-Gonzalez, Peter Pitalia, Rui Xu, Kyle Cranmer, David" }, { "sta...
hg2Q_O5b9w4	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	CURL: Contrastive Unsupervised Representations for Reinforcement Learning	[ "Science & Technology" ]	[ "deep learning", "machine learning", "rl", "reinforcement learning", "unsupervised", "contrast", "contrastive", "encoder", "self-supervised", "deep rl", "representation", "representation learning", "query", "key" ]	Contrastive Learning has been an established method in NLP and Image classification. The authors show that with relatively minor adjustments, CL can be used to augment and improve RL dramatically. Paper: https://arxiv.org/abs/2004.04136 Code: https://github.com/MishaLaskin/curl Abstract: We present CURL: Contrastive ...	Hi there! Today we're going to look at CURL, Contrastive Unsupervised Representations for Reinforcement Learning, by Aravind Srinivas, Michael Laskin and Pieter Abbeel. So this is a general framework for unsupervised representation learning for RL. So let's untangle the title a little bit. It is FOR reinforcement lear...	[ { "start": 0, "end": 7.5, "text": " Hi there! Today we're going to look at CURL, Contrastive Unsupervised Representations for Reinforcement Learning," }, { "start": 7.5, "end": 12.5, "text": " by Aravind Srinivas, Michael Laskin and Pieter Abbeel." }, { "start": 12.5, "end": ...
efPrtcLdcdM	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	GPT-4chan: This is the worst AI ever	[ "Science & Technology" ]	[]	#gpt4chan #4chan #ai GPT-4chan was trained on over 3 years of posts from 4chan's "politically incorrect" (/pol/) board. (and no, this is not GPT-4) EXTRA VIDEO HERE: https://www.youtube.com/watch?v=dQw4w9WgXcQ Website (try the model here): https://gpt-4chan.com Model (no longer available): https://huggingface.co/yk...	I trained an AI language model on three years worth of 4chan posts, I put the model into a chatbot. And in just a few days, it created 1000s of posts on the site as people slowly noticed that something strange is going on. I released the model, the code and I evaluated the model on a huge set of benchmarks. And it tur...	[ { "start": 0, "end": 5.4, "text": " I trained an AI language model on three years worth of 4chan posts, I put the model into" }, { "start": 5.4, "end": 6.4, "text": " a chatbot." }, { "start": 6.4, "end": 11.76, "text": " And in just a few days, it created 1000s of posts ...
NAJOZTNkhlI	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Language Models are Open Knowledge Graphs (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "nlp", "natural language processing", "bert", "gpt", "gpt2", "gpt-2", "gpt3", "gpt-3", "gpt 2", "gpt 3", "knowledge graph", "knowledge base", "language", ...	#ai #research #nlp Knowledge Graphs are structured databases that capture real-world entities and their relations to each other. KGs are usually built by human experts, which costs considerable amounts of time and money. This paper hypothesizes that language models, which have increased their performance dramatically ...	Hi there. Today we'll look at language models or open knowledge graphs by Cheng Wang Wang, Xiao Liu and Don Song. This paper on a high level proposes to construct knowledge graphs which is a structured object that's usually built by human, by experts, either fully manually or semi-manually with heavy human involvement...	[ { "start": 0, "end": 5.08, "text": " Hi there. Today we'll look at language models or open knowledge graphs by" }, { "start": 5.08, "end": 11.92, "text": " Cheng Wang Wang, Xiao Liu and Don Song. This paper on a high level proposes to" }, { "start": 11.92, "end": 16.76, "...
nXGHJTtFYRU	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Dynamic Routing Between Capsules	[ "Science & Technology" ]	[ "machine learning", "deep learning", "capsules", "capsule networks", "google brain", "hinton", "jeff hinton", "geoff hinton", "routing", "neural networks", "convolution", "convolutional neural networks", "deep neural networks", "cnns", "mnist", "multimnist", "disentanglement", "arc...	Geoff Hinton's next big idea! Capsule Networks are an alternative way of implementing neural networks by dividing each layer into capsules. Each capsule is responsible for detecting the presence and properties of one particular entity in the input sample. This information is then allocated dynamically to higher-level c...	Hi there! Today we're looking at dynamic routing between capsules by Sara Sabour, Nicholas Frost and Jeffrey Hinton of Google Brain. This paper is a bit older but it's made quite the impact at the time and so we'll go through it. I find this pretty hard paper to read and kind of understand because a lot of things are ...	[ { "start": 0, "end": 6, "text": " Hi there! Today we're looking at dynamic routing between capsules by Sara Sabour," }, { "start": 6, "end": 11.96, "text": " Nicholas Frost and Jeffrey Hinton of Google Brain. This paper is a bit older" }, { "start": 11.96, "end": 18.8, "t...
dPsXxLyqpfs	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	World Models	[ "Science & Technology" ]	[ "deep learning", "reinforcement learning", "deep reinforcement learning", "deep rl", "schmidhuber", "environment model", "imagination", "vae", "rnn", "lstm" ]	Authors: David Ha, Jürgen Schmidhuber Abstract: We explore building generative neural network models of popular reinforcement learning environments. Our world model can be trained quickly in an unsupervised manner to learn a compressed spatial and temporal representation of the environment. By using features extracted...	Hi, today we're looking at World Models by David Ha and Jürgen Schmidhuber. This is a paper that's concerned with reinforcement learning and especially with the problem of, say, you have an environment that you interact with and you kind of need to learn to act in it, but it could be, for example, very expensive to al...	[ { "start": 0, "end": 6, "text": " Hi, today we're looking at World Models by David Ha and Jürgen Schmidhuber." }, { "start": 6, "end": 13, "text": " This is a paper that's concerned with reinforcement learning and especially with the problem of," }, { "start": 13, "end": 20, ...
TrLrBL1U8z0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[ML News] GitHub Copilot - Copyright, GPL, Patents & more \| Brickit LEGO app \| Distill goes on break	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "copilot", "github copilot", "github copilot copyright", "github gpl", "github copilot gpl", "copilot copyright", "copilot gpl", "openai gpl", "openai copilot", "...	#copilot #copyright #gpl GitHub and OpenAI release Copilot, an AI-powered code autocomplete system that can generate entire functions, classes, and modules from mere definitions and docstrings. Copilot was trained on all public GitHub repositories, and this has a lot of people upset about questions on copyright, code ...	An open door. An open window. An open bottle. Open AI and GitHub invent Copilot and everyone freaks out about copyright. Welcome to ML News. Greg Brockman writes an AI pair programmer in your editor. It's powered by OpenAI Codex, a new AI system which can convert from natural language to code with increasing reliabili...	[ { "start": 0, "end": 2, "text": " An open door." }, { "start": 2, "end": 6, "text": " An open window." }, { "start": 6, "end": 10, "text": " An open bottle." }, { "start": 10, "end": 15, "text": " Open AI and GitHub invent Copilot and everyone freaks out a...
DiNzQP7kK-s	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Descending through a Crowded Valley -- Benchmarking Deep Learning Optimizers (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "optimization", "polyak", "nesterov", "benchmark", "cnn", "cifar", "mnist", "adam", "adagrad", "adadelta", "momentum", "sgd", "gradient", "learning rate",...	#ai #research #optimization Deep Learning famously gives rise to very complex, non-linear optimization problems that cannot be solved analytically. Therefore, the choice of a suitable optimization algorithm can often make or break the training of a Deep Neural Network. Yet, the literature is full with hundreds of diff...	Hi there, today we'll look at Descending Through a Crowded Valley, Benchmarking Deep Learning Optimizers by Robin M. Schmidt, Frank Schneider and Philipp Henning of the University of Tübingen. So this paper is an empirical investigation, a benchmark into optimization algorithms for deep learning. The short story of th...	[ { "start": 0, "end": 5.18, "text": " Hi there, today we'll look at Descending Through a Crowded Valley, Benchmarking Deep Learning" }, { "start": 5.18, "end": 11.76, "text": " Optimizers by Robin M. Schmidt, Frank Schneider and Philipp Henning of the University of Tübingen." }, { ...
H3Bhlan0mE0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Online Education - How I Make My Videos	[ "Science & Technology" ]	[ "deep learning", "machine learning", "online video", "university", "online", "create", "lecture" ]	Just a short overview of tools I use to make my videos. OneNote - https://www.onenote.com iSpring Free Cam - https://www.ispringsolutions.com/ispring-cam Shotcut - https://shotcut.org Slack - https://slack.com RocketChat - https://rocket.chat Zoom - https://zoom.us Jitsi - https://jitsi.org GDocs - https://www.google....	Hi there! So a lot of people have been asking me how I make these videos. And this is of course relevant now that everyone's work from home and all the schools are converted into online schools. All of a sudden a lot of people have to make these online education happen. And I think this style of video lends itself to ...	[ { "start": 0, "end": 5, "text": " Hi there! So a lot of people have been asking me how I make these videos." }, { "start": 5, "end": 13, "text": " And this is of course relevant now that everyone's work from home and all the schools are converted into online schools." }, { "start...
Ok44otx90D4	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Feature Visualization & The OpenAI microscope	[ "Science & Technology" ]	[ "deep learning", "machine learning", "imagenet", "visualization", "features", "intermediate", "hidden layers", "activations", "patterns", "openai", "google", "interactive", "explanation" ]	A closer look at the OpenAI microscope, a database of visualizations of the inner workings of ImageNet classifiers, along with an explanation of how to obtain these visualizations. https://distill.pub/2017/feature-visualization/ https://microscope.openai.com/models https://github.com/tensorflow/lucid Links: YouTube: ...	Hi there! Today we're going to take a look at the OpenAI microscope and this article on Distill called Feature Visualization. So the Feature Visualization article is by Chris Ola, Alexander Mortwintsev and Ludwig Schubert of the Google Brain team, while the OpenAI microscope is by OpenAI. So keep that in mind. These t...	[ { "start": 0, "end": 6.8, "text": " Hi there! Today we're going to take a look at the OpenAI microscope and this" }, { "start": 6.8, "end": 11.46, "text": " article on Distill called Feature Visualization. So the Feature" }, { "start": 11.46, "end": 17.72, "text": " Visua...
_EDr3ryrT_Y	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Typical Decoding for Natural Language Generation (Get more human-like outputs from language models!)	[ "Science & Technology" ]	[]	#deeplearning #nlp #sampling Modern language models like T5 or GPT-3 achieve remarkably low perplexities on both training and validation data, yet when sampling from their output distributions, the generated text often seems dull and uninteresting. Various workarounds have been proposed, such as top-k sampling and nuc...	Pay special attention to this paper. It is not a paper by Google or DeepMind or Meta or anything like this, yet I believe it is a really important paper. It discusses typical sampling, which is a new decoding strategy of how we sample from language models. We usually train language models with a maximum likelihood obj...	[ { "start": 0, "end": 7.04, "text": " Pay special attention to this paper. It is not a paper by Google or DeepMind or Meta or anything" }, { "start": 7.04, "end": 12.88, "text": " like this, yet I believe it is a really important paper. It discusses typical sampling, which is a" }, { ...
zWFkUGXjbdo	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[Rant] Can AI read your emotions? (No, but ...)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "ai face recognition", "face recognition", "face recognition emotion detection", "can ai read your mind", "can ai read your emotions", "ai emotion analysis", "ai analyz...	#facerecognition #emotiondetection #mindreading Face recognition has a bad rep in the ML community. While the technology continuously advances, so does the resistance against its applications, with good reasons: AI emotion analysis hints at a dystopian future where our lives are completely governed by algorithms. Howe...	We need to talk about your face or face recognition in general. A tweet has been making the rounds saying facial recognition is able to analyze in real time the emotions and feelings. Just that. And it showed a video of an apparent real-time system looking at people's faces and determining what their emotions are. Now...	[ { "start": 0, "end": 10.16, "text": " We need to talk about your face or face recognition in general. A tweet has been" }, { "start": 10.16, "end": 15.120000000000001, "text": " making the rounds saying facial recognition is able to analyze in real" }, { "start": 15.1200000000000...
FC-R4MlIqrc	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[ML News] Cedille French Language Model \| YOU Search Engine \| AI Finds Profitable MEME TOKENS	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "mlnews", "machine learning news", "ai news", "ml news", "cedille", "french language model", "gpt-j", "gpt j", "eleuther ai", "you", "you search", "you search...	#mlnews #cedille #wmt Only the greatest of news from the world of Machine Learning. OUTLINE: 0:00 - Sponsor: Weights & Biases 1:50 - Cedille - French Language Model 3:55 - Facebook AI Multilingual model wins WMT 5:50 - YOU private search engine 10:35 - DeepMind's Open-Source Arnheim 12:10 - Company sued for using AI ...	Hold on, this video is sponsored by weights and biases. Weights and biases is your one stop shop for all your machine learning needs. It will track your experiments with a single line of code will upload automatically all your logs, all your configurations, everything to your cloud, it will automatically grab all the ...	[ { "start": 0, "end": 9.68, "text": " Hold on, this video is sponsored by weights and biases. Weights and biases is your one" }, { "start": 9.68, "end": 14.96, "text": " stop shop for all your machine learning needs. It will track your experiments with a single" }, { "start": 14.9...
OUCwujwE7bA	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents (+Author)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "natural language processing", "training data", "deep learning tutorial", "nlp", "gpt3", "gpt 3", "codex", "openai codex", "large language models", "gpt 3 plannin...	#gpt3 #embodied #planning In this video: Paper explanation, followed by first author interview with Wenlong Huang. Large language models contain extraordinary amounts of world knowledge that can be queried in various ways. But their output format is largely uncontrollable. This paper investigates the VirtualHome envir...	Hello there, today we're looking at language models as zero-shot planners, extracting actionable knowledge for embodied agents. And I'm going to interview the first author Wenlong Huang in a few minutes. So first there's an explanation of the paper, 10-15 minutes or so, I'm going to try to keep to it. And then we jump...	[ { "start": 0, "end": 5.5200000000000005, "text": " Hello there, today we're looking at language models as zero-shot planners, extracting actionable" }, { "start": 5.5200000000000005, "end": 11.28, "text": " knowledge for embodied agents. And I'm going to interview the first author Wenlon...
hv3UO3G0Ofo	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Axial-DeepLab: Stand-Alone Axial-Attention for Panoptic Segmentation (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "google", "cnn", "resnet", "big bird", "bigbird", "attention", "attention mechanism", "attention for images", "transformer for images", "transformer", "bert", ...	#ai #machinelearning #attention Convolutional Neural Networks have dominated image processing for the last decade, but transformers are quickly replacing traditional models. This paper proposes a fully attentional model for images by combining learned Positional Embeddings with Axial Attention. This new model can comp...	Transformers are quickly coming for your favorite models. Yesterday they replaced LSTMs in NLP. They used to be good at NLP but we now have transformers. Think again. Today we're going to see that maybe in the near future transformers will replace convolutions in image processing. So this paper is a step in towards th...	[ { "start": 0, "end": 6, "text": " Transformers are quickly coming for your favorite models. Yesterday they replaced" }, { "start": 6, "end": 13.120000000000001, "text": " LSTMs in NLP. They used to be good at NLP but we now have transformers. Think again." }, { "start": 13.120000...
LB4B5FYvtdI	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Predictive Coding Approximates Backprop along Arbitrary Computation Graphs (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "backpropagation", "computation", "autograph", "tensorflow", "pytorch", "torch", "autodiff", "differentiation", "backprop", "biologically plausible", "neurons",...	#ai #biology #neuroscience Backpropagation is the workhorse of modern deep learning and a core component of most frameworks, but it has long been known that it is not biologically plausible, driving a divide between neuroscience and machine learning. This paper shows that Predictive Coding, a much more biologically pl...	Hi there, this is an LSTM cell or the computation graph of an LSTM cell. It is pretty hideous as you can see, but what I'm about to show you is even more hideous. This is the computation graph of the LSTM cell augmented with error units, evincing the connectivity scheme of the predictive coding algorithm. You may see ...	[ { "start": 0, "end": 7.76, "text": " Hi there, this is an LSTM cell or the computation graph of an LSTM cell. It is pretty hideous as you" }, { "start": 7.76, "end": 15.84, "text": " can see, but what I'm about to show you is even more hideous. This is the computation graph of the" }, ...
Pm93D8CVlY8	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	This A.I. creates infinite NFTs	[ "Science & Technology" ]	[]	#nft #gan #ai Today we build our own AI that can create as many bored apes as we want! Fungibility for everyone! Try the model here: https://huggingface.co/spaces/ykilcher/apes or here: https://ykilcher.com/apes Files & Models here: https://huggingface.co/ykilcher/apes/tree/main Code here: https://github.com/yk/apes-...	This ape does not exist. Neither does this one, this one, this, this, this or this. In fact, I've created all of them using an AI that I trained myself. And today I'm going to show you how it's done and what other cool things you can do with this. Hi there, my name is Yannick. Welcome to the channel. Today I'm going t...	[ { "start": 0, "end": 6.08, "text": " This ape does not exist. Neither does this one, this one, this, this, this or this. In fact," }, { "start": 6.08, "end": 10.88, "text": " I've created all of them using an AI that I trained myself. And today I'm going to show you" }, { "start"...
lmAj0SU_bW0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Axial Attention & MetNet: A Neural Weather Model for Precipitation Forecasting	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "google", "attention mechanism", "attention", "transformer", "rnn", "recurrent", "weather", "long-range", "layers", "convolutions", "cnns", "rain", "physics" ]	MetNet is a predictive neural network model for weather prediction. It uses axial attention to capture long-range dependencies. Axial attention decomposes attention layers over images into row-attention and column-attention in order to save memory and computation. https://ai.googleblog.com/2020/03/a-neural-weather-mod...	Hi there. So what you're looking at here is a weather forecast model. Specifically the very top row is a new weather forecast model called NetNet by Google Research. So the goal of weather prediction is pretty simple. You want to know what the weather is going to be in the future. Specifically here you want to know pr...	[ { "start": 0, "end": 8.52, "text": " Hi there. So what you're looking at here is a weather forecast model. Specifically the" }, { "start": 8.52, "end": 15.76, "text": " very top row is a new weather forecast model called NetNet by Google Research. So the goal" }, { "start": 15.76...
R5DiLFOMZrc	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	TransGAN: Two Transformers Can Make One Strong GAN (Machine Learning Research Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "neural networks", "ai", "artificial intelligence", "attention neural networks", "attention is all you need", "transformer gan", "transformer gans", "transformer generative adversarial network", "generative adversarial network", "attention mechan...	#transformer #gan #machinelearning Generative Adversarial Networks (GANs) hold the state-of-the-art when it comes to image generation. However, while the rest of computer vision is slowly taken over by transformers or other attention-based architectures, all working GANs to date contain some form of convolutional laye...	Hi there, today we'll look at TransGAN, two transformers can make one strong GAN, by Yifan Qian, Xu Yucheng and Cheng Yang Wang. So in this paper, the authors attempt to make a generative adversarial network, a GAN, out of only transformers. So far, attention or transformer-like things have been used in GANs, but they...	[ { "start": 0, "end": 7.5600000000000005, "text": " Hi there, today we'll look at TransGAN, two transformers can make one strong GAN, by Yifan" }, { "start": 7.5600000000000005, "end": 11.6, "text": " Qian, Xu Yucheng and Cheng Yang Wang." }, { "start": 11.6, "end": 17.76, ...
_N_nFzMtWkA	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Reinforcement Learning, Fast and Slow	[ "Science & Technology" ]	[ "machine learning", "reinforcement learning", "meta-learning", "deep rl", "deep reinforcement learning", "deep neural network", "atari", "alphago", "deepmind", "google", "td-gammon", "episodic memory", "inductive bias", "bias variance tradeoff" ]	Abstract: Deep reinforcement learning (RL) methods have driven impressive advances in artificial intelligence in recent years, exceeding human performance in domains ranging from Atari to Go to no-limit poker. This progress has drawn the attention of cognitive scientists interested in understanding human learning. Howe...	Hi there, today we're looking at reinforcement learning, fast and slow, by Matthew Botvinick, Sam Ritter, Jane X. Wang, Zeb Kurt-Nielsen, Charles Spondel and Demis Hassabis. These people are from Google DeepMind and this is a review of kind of a development in reinforcement learning, especially as it pertains to kind ...	[ { "start": 0, "end": 7, "text": " Hi there, today we're looking at reinforcement learning, fast and slow, by Matthew Botvinick," }, { "start": 7, "end": 17, "text": " Sam Ritter, Jane X. Wang, Zeb Kurt-Nielsen, Charles Spondel and Demis Hassabis." }, { "start": 17, "end": 24....
HYEzHX6-fIA	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Dynamics-Aware Unsupervised Discovery of Skills (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "rl", "deep rl", "control", "planning", "world model", "dads", "skills", "latent", "high level", "unsupervised", "tree search", "deep reinforcement learning",...	This RL framework can discover low-level skills all by itself without any reward. Even better, at test time it can compose its learned skills and reach a specified goal without any additional learning! Warning: Math-heavy! OUTLINE: 0:00 - Motivation 2:15 - High-Level Overview 3:20 - Model-Based vs Model-Free Reinforce...	Hi there! Take a look at this humanoid right here. It walks from one checkpoint to another checkpoint and then to the next checkpoint and so on. And that is its task. It gets reward from walking from checkpoint to checkpoint. Take a look at this ant. This is called the ant. It also walks from checkpoint to checkpoint....	[ { "start": 0, "end": 8.44, "text": " Hi there! Take a look at this humanoid right here. It walks from one checkpoint to another" }, { "start": 8.44, "end": 13.76, "text": " checkpoint and then to the next checkpoint and so on. And that is its task. It gets reward" }, { "start": 1...
XdpF9ZixIbI	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Can we Contain Covid-19 without Locking-down the Economy?	[ "Science & Technology" ]	[ "machine learning", "epidemiology", "worst case", "statistics", "hypothesis test", "covid", "corona", "coronavirus" ]	My thoughts on the let-the-young-get-infected argument. https://medium.com/amnon-shashua/can-we-contain-covid-19-without-locking-down-the-economy-2a134a71873f Abstract: In this article, we present an analysis of a risk-based selective quarantine model where the population is divided into low and high-risk groups. The...	Can we contain COVID-19 without locking down the economy? This is a question and I do care about this article because Shai Shalef-Schwarz is one of the bigger names in machine learning theory. So it was interesting for me to see what he and his collaborator here had to say about the kind of outbreak and the strategy t...	[ { "start": 0, "end": 6, "text": " Can we contain COVID-19 without locking down the economy?" }, { "start": 6, "end": 11, "text": " This is a question and I do care about this article because" }, { "start": 11, "end": 16, "text": " Shai Shalef-Schwarz is one of the bigger ...
Ru23eWAQ6_E	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Do As I Can, Not As I Say: Grounding Language in Robotic Affordances (SayCan - Paper Explained)	[ "Science & Technology" ]	[]	#saycan #robots #ai Large Language Models are excellent at generating plausible plans in response to real-world problems, but without interacting with the environment, they have no abilities to estimate which of these plans are feasible or appropriate. SayCan combines the semantic capabilities of language models with ...	Hi there, check out this video. So there's a Coke can and there's a spill, a Coke spill. So the instructor here says, I spilled my Coke on the table. How would you throw it away and bring me something to help clean? So the robot here forms a plan as it goes about it. First, it says I would find a Coke can. Then second...	[ { "start": 0, "end": 7.2, "text": " Hi there, check out this video. So there's a Coke can and there's a spill, a Coke spill." }, { "start": 7.2, "end": 13.36, "text": " So the instructor here says, I spilled my Coke on the table. How would you throw it away and" }, { "start": 13....
RJwPN4qNi_Y	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[ML News] Google's 540B PaLM Language Model & OpenAI's DALL-E 2 Text-to-Image Revolution	[ "Science & Technology" ]	[]	#mlnews #palm #dalle2 Google releases PaLM and OpenAI releases DALL-E 2 (and more news). Sponsor: Weights & BIases Start here: https://wandb.me/yannic Thumbnail credit: DALL-E 2 via Sam Altman OUTLINE 0:00 - Street interview w/ random stranger 2:25 - Intro 2:50 - PaLM - Google's 540B Pathways Language Model 7:50 - ...	So I was wondering what happens if you just ask some random people on the street about this paper and... Actually... Sir, sir, excuse me sir. Hi, how are you doing? I was wondering what do you think about this new paper by Google, this Palm paper, however they call it. The Palm paper? You mean the latest large languag...	[ { "start": 0, "end": 6, "text": " So I was wondering what happens if you just ask some random people on the street about this paper and..." }, { "start": 6, "end": 7, "text": " Actually..." }, { "start": 7, "end": 10, "text": " Sir, sir, excuse me sir." }, { "star...
IiBFqnNu7A8	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Planning to Explore via Self-Supervised World Models (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "rl", "deep rl", "deep reinforcement learning", "novelty", "curiosity", "intrinsic reward", "dreamer", "planet", "control", "walker", "run forward", "imaginar...	What can an agent do without any reward? Explore the world! While many formulations of intrinsic rewards exist (Curiosity, Novelty, etc.), they all look back in time to learn. Plan2Explore is the first model that uses planning in a learned imaginary latent world model to seek out states where it is uncertain about what...	Hi there! Today we're looking at planning to explore via self-supervised world models by Ramanan Sekar, Ole Rybkin, Kostas Danilidis, Peter Abil, Danijar Hafner and Depak Patak. So this is a paper that concerns reinforcement learning and specifically self-supervised reinforcement learning. So what do they mean? Here's...	[ { "start": 0, "end": 6.640000000000001, "text": " Hi there! Today we're looking at planning to explore via self-supervised world models" }, { "start": 6.640000000000001, "end": 17.04, "text": " by Ramanan Sekar, Ole Rybkin, Kostas Danilidis, Peter Abil, Danijar Hafner and Depak Patak." ...
F5mxzvgl_oU	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	S.H.E. - Search. Human. Equalizer.	[ "Science & Technology" ]	[ "pantene", "search", "google", "bias", "machine learning", "artificial intelligence", "search engine", "ranking", "equality", "diversity" ]	Short opinion on Pantene's tool to de-bias Google search results. https://www.apnews.com/Business%20Wire/c53a0e8f5fe04bf68e8311f214c806cf https://shetransforms.us/	Hi everyone, just a quick more of a news update in the AI world. Which is the following. Pantene launches S.H.E. The Search Human Equalizer to shine a light on bias in search. So Pantene, the kind of cosmetic corporation, launches this thing which is supposed to correct your search. And it's introduced here in this Yo...	[ { "start": 0, "end": 7.12, "text": " Hi everyone, just a quick more of a news update in the AI world." }, { "start": 7.12, "end": 8.96, "text": " Which is the following." }, { "start": 8.96, "end": 11.120000000000001, "text": " Pantene launches S.H.E." }, { "start...
NJCLUzkn-sA	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	EfficientZero: Mastering Atari Games with Limited Data (Machine Learning Research Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "muzero", "alphazero", "berkeley", "pieter abbeel", "dreamer", "dreamerv2", "atari", "reinforcement learning", "deep reinforcement learning", "world model", "le...	#efficientzero #muzero #atari Reinforcement Learning methods are notoriously data-hungry. Notably, MuZero learns a latent world model just from scalar feedback of reward- and policy-predictions, and therefore relies on scale to perform well. However, most RL algorithms fail when presented with very little data. Effici...	Hi there, today we're going to look at Mastering Atari Games with Limited Data by Waziru Yeh, Shahuwa Liu, Tanahar Kuretach, Pietra Biel and Yang Gao. This paper presents the Efficient Zero model, which is a model that can do reinforcement learning with severely limited data. So the paper tackles the Atari 100k benchm...	[ { "start": 0, "end": 5.84, "text": " Hi there, today we're going to look at Mastering Atari Games with Limited Data by Waziru Yeh," }, { "start": 5.84, "end": 13.76, "text": " Shahuwa Liu, Tanahar Kuretach, Pietra Biel and Yang Gao. This paper presents the Efficient Zero" }, { "s...
NEkriziVYXo	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[ML News] DeepMind does Nowcasting \| The Guardian's shady reporting \| AI finishes Beethoven's 10th	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "nowcasting", "mlnews", "deepmind", "weather prediction", "ai weather", "short term weather", "rain prediction", "rain when", "ai nowcasting", "ml weather predict...	#deepmind #nowcasting #machinelearning Your holy update on what's new in the Machine Learning world. OUTLINE: 0:00 - Intro 0:30 - DeepMind tackles Nowcasting 3:30 - The Guardian's shady reporting on TruthfulQA 6:15 - Stochastic training not necessary for generalization 7:35 - Google AI's efficient partitioning of roa...	Cut my hair, but not the beard. I have a giant cold sore here. That just looks weird without the beard. I was just gonna wait. Well, we'll... um, yeah. Intro. DeepMind can predict rain better than anyone else. The Guardian is not so really truthful about truthful language models. And an AI finishes Beethoven's 10th sy...	[ { "start": 0, "end": 4.4, "text": " Cut my hair, but not the beard. I have a giant cold sore here." }, { "start": 4.4, "end": 8.08, "text": " That just looks weird without the beard. I was just gonna wait." }, { "start": 8.08, "end": 11.040000000000001, "text": " Well, we...
yexR53My2O4	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[Classic] Word2Vec: Distributed Representations of Words and Phrases and their Compositionality	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "jeff dean", "mikolov", "word2vec", "word vectors", "word representations", "nlp", "natural language processing", "sentiment classification", "king", "queen", "...	#ai #research #word2vec Word vectors have been one of the most influential techniques in modern NLP to date. This paper describes Word2Vec, which the most popular technique to obtain word vectors. The paper introduces the negative sampling technique as an approximation to noise contrastive estimation and shows that th...	Hi there, today we'll look at distributed representations of words and phrases and their compositionality by Thomas Mikolov, Ilya Sotskyver, Kai Chen, Greg Corrado and Jeffrey Dean. This is another historical paper, it's one of three papers, it's the middle one that introduces the original Word2vec algorithm. And as y...	[ { "start": 0, "end": 5.5200000000000005, "text": " Hi there, today we'll look at distributed representations of words and phrases and their" }, { "start": 5.5200000000000005, "end": 12.56, "text": " compositionality by Thomas Mikolov, Ilya Sotskyver, Kai Chen, Greg Corrado and Jeffrey De...
utuz7wBGjKM	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[News] OpenAI Model Generates Python Code	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "microsoft", "openai", "msbuild", "build", "code", "gpt2", "language model", "completion", "intellisense", "intellicode", "vscode", "github", "python", "c...	This code completion engine can write an entire function from just the name! OpenAI demonstrates what happens when you learn a language model on thousands of GitHub Python repositories. Source Clip: https://youtu.be/fZSFNUT6iY8 Full Video: https://www.pscp.tv/Microsoft/1OyKAYWPRrWKb Kite: https://kite.com/ TabNine: ht...	Hi there. So I saw this and probably many of you have seen this. OpenAI was demonstrating at MSBuild basically a GPT-2 language model but trained not on language but on code, on Python code, open source code from GitHub. And so the idea is that the model learns to produce code. And we'll just have a short look at the ...	[ { "start": 0, "end": 8, "text": " Hi there. So I saw this and probably many of you have seen this. OpenAI was demonstrating at MSBuild" }, { "start": 8, "end": 14.88, "text": " basically a GPT-2 language model but trained not on language but on code, on Python code," }, { "start"...
p3sAF3gVMMA	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Deep Learning for Symbolic Mathematics	[ "Science & Technology" ]	[ "deep learning", "machine learning", "nlp", "natural language processing", "machine translation", "arxiv", "attention mechanism", "attention", "transformer", "rnn", "recurrent", "seq2seq", "facebook", "fair", "research", "math", "integral", "ode" ]	This model solves integrals and ODEs by doing seq2seq! https://arxiv.org/abs/1912.01412 https://ai.facebook.com/blog/using-neural-networks-to-solve-advanced-mathematics-equations/ Abstract: Neural networks have a reputation for being better at solving statistical or approximate problems than at performing calculation...	Hi there! Can you solve this? Neither can I. But Wolfram Alpha can. So this is the thing that probably I have most to thank for for passing university, especially the math classes in it. If you don't know Wolfram Alpha, it is an engine from the creators of Mathematica, but it is online. It can do symbolic math. So it ...	[ { "start": 0, "end": 16, "text": " Hi there! Can you solve this? Neither can I. But Wolfram Alpha can. So this is the thing that probably I have most to thank for for passing university, especially the math classes in it." }, { "start": 16, "end": 33, "text": " If you don't know Wolfram ...
BBp0tHcirtQ	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	git for research basics: fundamentals, commits, branches, merging	[ "Science & Technology" ]	[ "git", "research", "commit", "merge", "conflict" ]	Don't watch this if you already know how to solve a merge conflict :)	Hi there. Today we're taking a look at Git, especially Git as it is used maybe in research collaborations. So Git is like a tool to collaborate, but when you research, like when you work on a paper together with other people, you won't use a lot of the features that Git offers and that are usually described by Git. So...	[ { "start": 0, "end": 9, "text": " Hi there. Today we're taking a look at Git, especially Git as it is used maybe in research collaborations." }, { "start": 9, "end": 19, "text": " So Git is like a tool to collaborate, but when you research, like when you work on a paper together with oth...
AU30czb4iQA	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Imputer: Sequence Modelling via Imputation and Dynamic Programming	[ "Science & Technology" ]	[ "deep learning", "machine learning", "nlp", "natural language processing", "machine translation", "arxiv", "google", "attention mechanism", "attention", "transformer", "seq2seq", "autoregressive", "independence", "decoding" ]	The imputer is a sequence-to-sequence model that strikes a balance between fully autoregressive models with long inference times and fully non-autoregressive models with fast inference. The imputer achieves constant decoding time independent of sequence length by exploiting dynamic programming. https://arxiv.org/abs/2...	Hi there! Today we're looking at the imputer sequence modeling via imputation and dynamic programming by William Chan, Chitwan Sariah, Jeffrey Hinton, Mohamed Nourouzi and Navdeep Jaitley. So this is a model to perform sequence-to-sequence tasks. Now sequence-to-sequence tasks are very very common in NLP, but in this ...	[ { "start": 0, "end": 6.12, "text": " Hi there! Today we're looking at the imputer sequence modeling via imputation" }, { "start": 6.12, "end": 12.72, "text": " and dynamic programming by William Chan, Chitwan Sariah, Jeffrey Hinton, Mohamed" }, { "start": 12.72, "end": 18.96,...
WVPE62Gk3EM	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Big Bird: Transformers for Longer Sequences (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "google", "google research", "bigbird", "big bird", "bert", "attention", "attention is all you need", "longformer", "random attention", "quadratic attention", "...	#ai #nlp #attention The quadratic resource requirements of the attention mechanism are the main roadblock in scaling up transformers to long sequences. This paper replaces the full quadratic attention mechanism by a combination of random attention, window attention, and global attention. Not only does this allow the p...	Hi there, today we'll look at Big Bird Transformers for Longer Sequences by Manil Zahir and Gurugar Uganesh et al. of Google Research. So this paper on a high level proposes to replace the quadratic attention mechanism in transformers by a mix of random attention, windowed attention, and selective global attention, th...	[ { "start": 0, "end": 6.88, "text": " Hi there, today we'll look at Big Bird Transformers for Longer Sequences by Manil Zahir and Gurugar" }, { "start": 6.88, "end": 9.84, "text": " Uganesh et al. of Google Research." }, { "start": 9.84, "end": 14.5, "text": " So this pape...
O9kFX33nUcU	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	On the Measure of Intelligence by François Chollet - Part 4: The ARC Challenge (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "chollet", "keras", "google", "francois", "intelligence", "iq", "iq test", "deep neural networks", "prior", "skill", "performance", "measurement", "measure"...	In this part, we look at the ARC challenge as a proposed test of machine intelligence. The dataset features 1000 tasks that test rapid generalization based on human core knowledge priors, such as object-ness, symmetry, and navigation. OUTLINE: 0:00 - Intro 0:55 - What is ARC? 6:30 - The Goals of ARC 10:40 - Assumed Pr...	Hi there and welcome to the last part of On the Measure of Intelligence by François Chollet. This last part concerns the ARC challenge that Chollet has proposed, or the ARC dataset, which stands for the Abstraction and Reasoning Corpus. And we're just quickly going over the dataset, look how it's built and discuss wha...	[ { "start": 0, "end": 9.040000000000001, "text": " Hi there and welcome to the last part of On the Measure of Intelligence by François Chollet." }, { "start": 9.040000000000001, "end": 16, "text": " This last part concerns the ARC challenge that Chollet has proposed, or the ARC dataset," ...
bFn2xcGi1TQ	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Faster Neural Network Training with Data Echoing (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "google", "brain", "pipeline", "bottleneck", "speed", "gpu", "tpu", "idle", "network", "distributed", "preprocessing", "augmentation" ]	CPUs are often bottlenecks in Machine Learning pipelines. Data fetching, loading, preprocessing and augmentation can be slow to a point where the GPUs are mostly idle. Data Echoing is a technique to re-use data that is already in the pipeline to reclaim this idle time and keep the GPUs busy at all times. https://arxiv...	Hi there! Today we're looking at faster neural network training with data echoing by Damme Choi, Alexander Passo, Christopher J. Shalu and George E. Dahl. So on a high level this paper basically says you should repeat data that's already in memory in order to speed up the entire process of neural network training. And...	[ { "start": 0, "end": 5.12, "text": " Hi there! Today we're looking at faster neural network training with data" }, { "start": 5.12, "end": 11.9, "text": " echoing by Damme Choi, Alexander Passo, Christopher J. Shalu and George E. Dahl." }, { "start": 11.9, "end": 17.34, "...
ifBI2jTaAEo	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Celebrating 100k Subscribers! (w/ Channel Statistics)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper" ]	#yannickilcher #machinelearning #100k OUTLINE: 0:00 - 100k! 1:00 - Announcements & Thanks 3:55 - Channel Statistics Links: TabNine Code Completion (Referral): http://bit.ly/tabnine-yannick YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://discord.gg/4H8xxDF BitChu...	Yay! 100k! Nice! Big celebration, we have just reached 100,000 subscribers. Now truth be told as of recording of this videos, we actually don't have 100,000 subscribers yet. There's like 156 missing. So all I have to do is not get cancelled in the next two days or so. And this is harder than it seems. But I've managed...	[ { "start": 0, "end": 12.280000000000001, "text": " Yay! 100k! Nice! Big celebration, we have just reached 100,000 subscribers. Now truth" }, { "start": 12.280000000000001, "end": 17.96, "text": " be told as of recording of this videos, we actually don't have 100,000 subscribers yet." }...
RXwZKzczkF8	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[ML News] AI Threatens Biological Arms Race	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "gtc", "gtc22", "nvidia", "jensen huang", "3090", "rtx 3090", "ithaca", "deepmind", "deep mind", "deepmind greek text", "deepmind ithaca", "ml news", "mlnew...	#mlnews #gtc22 #ithaca GTC Registration Link: https://ykilcher.com/gtc Your regular updates on what's going on in the ML world! OUTLINE: 0:00 - Intro 0:20 - Register to Nvidia GTC and win a 3090! 4:15 - DeepMind's Ithaca deciphers Lost Ancient Texts 6:45 - Drug discovery model turns toxic 10:00 - Gary Marcus: Deep Le...	DeepMind uses deep learning to restore ancient Greek texts. A drug discovery system has been abused to create thousands and thousands of super toxic compounds. And Gary Marcus claims deep learning is hitting a wall. Welcome to ML News. It's Monday. GTC conference goes into its next iteration. Now GTC is a company conf...	[ { "start": 0, "end": 6.5600000000000005, "text": " DeepMind uses deep learning to restore ancient Greek texts. A drug discovery system has been" }, { "start": 6.5600000000000005, "end": 12.32, "text": " abused to create thousands and thousands of super toxic compounds. And Gary Marcus cl...
pH2jZun8MoY	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Involution: Inverting the Inherence of Convolution for Visual Recognition (Research Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "what is deep learning", "deep learning tutorial", "introduction to deep learning", "computer vision", "convolutional neural network", "convolutions alternative", "cnn ...	#involution #computervision #attention Convolutional Neural Networks (CNNs) have dominated computer vision for almost a decade by applying two fundamental principles: Spatial agnosticism and channel-specific computations. Involution aims to invert these principles and presents a spatial-specific computation, which is ...	Hello there! Today we're looking at involution, inverting the inheritance of convolution for visual recognition by a number of researchers of the Hong Kong University of Science and Technology, ByteDance AI lab and Peking University. In this paper on a high level the researchers try to replace the good old convolution...	[ { "start": 0, "end": 5.44, "text": " Hello there! Today we're looking at involution, inverting the inheritance of" }, { "start": 5.44, "end": 9.78, "text": " convolution for visual recognition by a number of researchers of the Hong Kong" }, { "start": 9.78, "end": 14.76, ...
smxwT82o40Y	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Active Dendrites avoid catastrophic forgetting - Interview with the Authors	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "active dendrites", "neurons dendrites", "biological deep learning", "deep learning biology", "numenta", "numenta research", "numenta deep learning", "dendrites deep ...	#multitasklearning #biology #neuralnetworks This is an interview with the paper's authors: Abhiram Iyer, Karan Grewal, and Akash Velu! Paper Review Video: https://youtu.be/O_dJ31T01i8 Check out Zak's course on Graph Neural Networks (discount with this link): https://www.graphneuralnets.com/p/introduction-to-gnns?coup...	Hello, this is an interview with the authors of the paper on active dendrites. Now, if you haven't seen it, I've made a comprehensive paper review video on this paper and I released that yesterday. If you watch this video as it comes out, which obviously you do today, I'm going to interview the authors and we've all s...	[ { "start": 0, "end": 10.64, "text": " Hello, this is an interview with the authors of the paper on active dendrites. Now, if" }, { "start": 10.64, "end": 16.94, "text": " you haven't seen it, I've made a comprehensive paper review video on this paper and I released" }, { "start":...
rd3R_G6_UfY	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Full Self-Driving is HARD! Analyzing Elon Musk re: Tesla Autopilot on Lex Fridman's Podcast	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "lex fridman", "elon musk", "elon", "musk", "tesla fsd", "when will fsd ship", "when will fsd be ready", "tesla fsd release", "tesla fsd release date", "how does ...	#tesla #fsd #elon Watch the original podcast: https://www.youtube.com/watch?v=DxREm3s1scA An analysis of Elon's appearance on Lex Fridman. Very interesting conversation and a good overview of past, current, and future versions of Tesla's Autopilot system. OUTLINE: 0:00 - Intro 0:40 - Tesla Autopilot: How hard is it?...	Hey, how's everyone doing today? We're going to analyze Elon Musk's appearance on the Lex Friedman podcast. Specifically, we're going to look at the part where Elon talks about the Tesla autopilot and to a certain degree, also the Tesla bot. We've previously analyzed the talk by Andrej Karpati about what kind of archi...	[ { "start": 0, "end": 1.32, "text": " Hey, how's everyone doing today?" }, { "start": 1.32, "end": 6.16, "text": " We're going to analyze Elon Musk's appearance on the Lex Friedman podcast." }, { "start": 6.24, "end": 10.120000000000001, "text": " Specifically, we're going...
-_2AF9Lhweo	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Linformer: Self-Attention with Linear Complexity (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "facebook", "linear", "quadratic", "transformer", "attention", "self-attention", "multi-head attention", "t2t", "vasvani", "bert", "devlin", "roberta", "glu...	Transformers are notoriously resource-intensive because their self-attention mechanism requires a squared number of memory and computations in the length of the input sequence. The Linformer Model gets around that by using the fact that often, the actual information in the attention matrix is of lower rank and can be a...	Hi there! Today we're going to look at Linformer self-attention with linear complexity by Sinon Wang, Belinda Li, Madian Kabsa, Han Feng and Hao Ma of Facebook AI. So on a high level this paper observes that often the way we build transformers the self-attention matrix is low rank and can be approximated by first proj...	[ { "start": 0, "end": 4.96, "text": " Hi there! Today we're going to look at Linformer self-attention with linear" }, { "start": 4.96, "end": 11.84, "text": " complexity by Sinon Wang, Belinda Li, Madian Kabsa, Han Feng and Hao Ma of" }, { "start": 11.84, "end": 17.94, "te...
G3pOvrKkFuk	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[Code] PyTorch sentiment classifier from scratch with Huggingface NLP Library (Full Tutorial)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "code", "pytorch", "bert", "pretrained", "lightning", "live", "tutorial", "pip", "nlp", "transformers", "tokenizers", "sequence", "sentiment", "imdb", "...	Huggingface released its newest library called NLP, which gives you easy access to almost any NLP dataset and metric in one convenient interface. We will combine this with a BERT model from Huggingface's Transformers library to build a sentiment classifier for IMDB. OUTLINE: 0:00 - Intro 1:30 - Boilerplate 3:20 - PyTo...	How did it do? So Hugging Face just released this NLP library right here and this is pretty cool because it allows you access to about a hundred NLP data sets and ten evaluation metrics pre-packaged. So knowing Hugging Face this is going to be a breeze to work with. So what I thought we would do is we would try to use...	[ { "start": 0, "end": 6.96, "text": " How did it do? So Hugging Face just released this NLP library right here and" }, { "start": 6.96, "end": 13.48, "text": " this is pretty cool because it allows you access to about a hundred NLP data" }, { "start": 13.48, "end": 19.3, "...
CA8JPbJ75tY	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	CornerNet: Detecting Objects as Paired Keypoints (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "corner", "top left", "bottom right", "corners", "cv", "computer vision", "vision", "object detection", "detr", "bounding box", "center", "anchor", "pooling...	Many object detectors focus on locating the center of the object they want to find. However, this leaves them with the secondary problem of determining the specifications of the bounding box, leading to undesirable solutions like anchor boxes. This paper directly detects the top left and the bottom right corners of obj...	Hello there! Today we're looking at corner net detecting objects as paired key points by Hylah and Jia Ding. So on a high level this paper detects objects in images. Let's say this is an image and here's a chair. You know you have your chair. And the way you detect the chair for this paper is going to be you detect th...	[ { "start": 0, "end": 6.08, "text": " Hello there! Today we're looking at corner net detecting objects as paired key points by" }, { "start": 6.08, "end": 15.280000000000001, "text": " Hylah and Jia Ding. So on a high level this paper detects objects in images. Let's say this is an" }, ...
x6T1zMSE4Ts	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	NVAE: A Deep Hierarchical Variational Autoencoder (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "gan", "vae", "kl", "elbo", "autoencoder", "variational", "latent", "sampling", "hierarchical", "scales", "faces", "mnist", "cifar10", "swish", "batch n...	VAEs have been traditionally hard to train at high resolutions and unstable when going deep with many layers. In addition, VAE samples are often more blurry and less crisp than those from GANs. This paper details all the engineering choices necessary to successfully train a deep hierarchical VAE that exhibits global co...	Alright, hi there. Have a look at these faces right here. So you're probably used by now to seeing computer-generated faces of really high quality, but probably you're used to seeing these faces coming from a generative adversarial network. However, these faces right here are from a variational autoencoder. Now, varia...	[ { "start": 0, "end": 5.6000000000000005, "text": " Alright, hi there. Have a look at these faces right here. So you're probably used by now to seeing" }, { "start": 5.6000000000000005, "end": 11.200000000000001, "text": " computer-generated faces of really high quality, but probably you'...
F5aaXrIMWyU	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	The AI Economist: Improving Equality and Productivity with AI-Driven Tax Policies (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "reinforcement learning", "society", "gini index", "welfare", "taxes", "brackets", "progressive", "regressive", "us", "poor", "rich", "equality", "redistribution", "outer loop", "world", "resources", "labor", "trade", "neural networks", "ppo" ]	Hail the AI Tax Collector! This very visual framework has RL Agents maximize their coins in a tiny world through collecting, building and trading. But at the same time, the government is also an AI trying to maximize social welfare via taxes. What emerges is very interesting. Paper: https://arxiv.org/abs/2004.13332 Bl...	Alright, today we're going to find out why AI is much better at governing people, why poor people really should pay more taxes, and how Donald Trump is just a normal human. Alright, we'll dive into it. We're looking at the AI Economist by Salesforce Research. Now Salesforce Research has kind of created a simulated wor...	[ { "start": 0, "end": 5.76, "text": " Alright, today we're going to find out why AI is much better at governing people, why" }, { "start": 5.76, "end": 12.8, "text": " poor people really should pay more taxes, and how Donald Trump is just a normal human." }, { "start": 12.8, "...
kl3aBni87jg	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	First Author Interview: AI & formal math (Formal Mathematics Statement Curriculum Learning)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "openai", "formal math", "ai math", "ai math prover", "machine learning for math", "ml math", "artificial intelligence math", "ai mathematics", "automated proof sea...	#openai #math #imo This is an interview with Stanislas Polu, research engineer at OpenAI and first author of the paper "Formal Mathematics Statement Curriculum Learning". Watch the paper review here: https://youtu.be/lvYVuOmUVs8 OUTLINE: 0:00 - Intro 2:00 - How do you explain the big public reaction? 4:00 - What's th...	Hello there, this is an interview with the first author of the paper, Formal Mathematics Statement Curriculum Learning, in which an automated system was able to solve two problems of the International Mathematics Olympiad. Now, this is an unprecedented level of skill in formal mathematics for an AI system. The system ...	[ { "start": 0, "end": 9.32, "text": " Hello there, this is an interview with the first author of the paper, Formal Mathematics" }, { "start": 9.32, "end": 15.540000000000001, "text": " Statement Curriculum Learning, in which an automated system was able to solve two problems" }, { ...
2h4tRsQzipQ	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Autoregressive Diffusion Models (Machine Learning Research Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "diffusion models", "autoregressive models", "generative models", "nlp", "natural language processing", "gpt", "image-gpt", "gpt-3", "gpt-2", "order agnostic", ...	#machinelearning #ardm #generativemodels Diffusion models have made large advances in recent months as a new type of generative models. This paper introduces Autoregressive Diffusion Models (ARDMs), which are a mix between autoregressive generative models and diffusion models. ARDMs are trained to be agnostic to the o...	Hi there! Today we'll look at autoregressive diffusion models by Emil Hageboom and others of Google research. This paper on a high level proposes a new type of autoregressive model, specifically one where variables can be decoded in arbitrary orders. This is akin to the new types of diffusion models that have been use...	[ { "start": 0, "end": 6.24, "text": " Hi there! Today we'll look at autoregressive diffusion models by Emil Hageboom and others" }, { "start": 6.24, "end": 12.92, "text": " of Google research. This paper on a high level proposes a new type of autoregressive model," }, { "start": 1...
ZfDZRX3WiJg	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	VirTex: Learning Visual Representations from Textual Annotations (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "cnn", "visual", "resnet", "caption", "nlp", "transformer", "vasvani", "attention", "text", "coco", "imagenet", "convolutional neural network", "adaptation"...	Pre-training a CNN backbone for visual transfer learning has recently seen a big push into the direction of incorporating more data, at the cost of less supervision. This paper investigates the opposite: Visual transfer learning by pre-training from very few, but very high-quality samples on an image captioning task. ...	Hi there! Today we're looking at Vertex Learning Visual Representations from Textual Annotations by Karen Desai and Justin Johnson of the University of Michigan. So this paper at its core is pretty simple. On a high level it proposes to take the task of image captioning, which is where you're given an image and you're...	[ { "start": 0, "end": 6.08, "text": " Hi there! Today we're looking at Vertex Learning Visual Representations from Textual Annotations" }, { "start": 6.08, "end": 13, "text": " by Karen Desai and Justin Johnson of the University of Michigan. So this paper at its core is pretty" }, { ...
_7xpGve9QEE	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	The Future of AI is Self-Organizing and Self-Assembling (w/ Prof. Sebastian Risi)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "sebastian risi", "copenhagen", "minecraft ai", "self-assembly", "self assembly", "nanobots", "swarm bots", "swarm ai", "evolution ai", "evolutionary methods", ...	#ai #selforganization #emergence Read Sebastian's article here: https://sebastianrisi.com/self_assembling_ai/ OUTLINE: 0:00 - Introduction 2:25 - Start of Interview 4:00 - The intelligence of swarms 9:15 - The game of life & neural cellular automata 14:10 - What's missing from neural CAs? 17:20 - How does local compu...	Hey there, today I'm talking to Sebastian Riese, who is the director of the creative AI lab and the co director of the robotics, evolution and art lab at the IT University of Copenhagen. He's also the co founder of a company called model AI that uses AI for various aspects of game development. Specifically today, we'r...	[ { "start": 0, "end": 3.9, "text": " Hey there, today I'm talking to Sebastian Riese, who is the director of the creative" }, { "start": 3.9, "end": 9.040000000000001, "text": " AI lab and the co director of the robotics, evolution and art lab at the IT University" }, { "start": 9...
qSArFEIoSbo	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	RepNet: Counting Out Time - Class Agnostic Video Repetition Counting in the Wild (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "vision", "counting", "self-similarity", "temporal", "frames", "video", "repeating", "lines", "transformer", "attention", "cnn", "convolutional neural network...	Counting repeated actions in a video is one of the easiest tasks for humans, yet remains incredibly hard for machines. RepNet achieves state-of-the-art by creating an information bottleneck in the form of a temporal self-similarity matrix, relating video frames to each other in a way that forces the model to surface th...	Hi there! Check out these videos on the top. Each one kind of contains a repeating action. So on the left you see someone doing jumping jacks in a fairly regular pattern. In the middle it gets a bit more difficult because what you see is a tennis ball bouncing and it bounces faster and faster and faster as time goes o...	[ { "start": 0, "end": 7.68, "text": " Hi there! Check out these videos on the top. Each one kind of contains a repeating action." }, { "start": 7.68, "end": 12.92, "text": " So on the left you see someone doing jumping jacks in a fairly regular pattern. In the" }, { "start": 12.92...
_Z9ZP1eiKsI	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Curiosity-driven Exploration by Self-supervised Prediction	[ "Science & Technology" ]	[]	https://arxiv.org/abs/1705.05363 Authors: Deepak Pathak, Pulkit Agrawal, Alexei A. Efros, Trevor Darrell Abstract: In many real-world scenarios, rewards extrinsic to the agent are extremely sparse, or absent altogether. In such cases, curiosity can serve as an intrinsic reward signal to enable the agent to explore it...	Hi there! Today we're going to look at this paper, Curiosity-Driven Exploration by Self-Supervised Prediction. It's a relatively short idea, so it shouldn't take too long. So the fundamental idea of the paper is to tackle the reward sparseness problem reinforcement learning. For example, if you have a Super Mario game...	[ { "start": 0, "end": 8, "text": " Hi there! Today we're going to look at this paper, Curiosity-Driven Exploration by Self-Supervised" }, { "start": 8, "end": 14.84, "text": " Prediction. It's a relatively short idea, so it shouldn't take too long. So the fundamental" }, { "start"...
19Q-vMd9bYg	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Inconsistency in Conference Peer Review: Revisiting the 2014 NeurIPS Experiment (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "neurips", "nips", "nips experiment", "peer reviw", "conference review", "reviewer", "machine learning reviewer", "ml conference review", "subjectivity in peer revi...	#neurips #peerreview #nips The peer-review system at Machine Learning conferences has come under much criticism over the last years. One major driver was the infamous 2014 NeurIPS experiment, where a subset of papers were given to two different sets of reviewers. This experiment showed that only about half of all acce...	Hi there, today we'll look at inconsistency in conference peer review, revisiting the 2014 NeurIPS experiment by Corina Cortes and Neil D. Lawrence, which were actually the chairs of the 2014 NeurIPS conference. So they are going to have access to some data that the rest of us sadly don't have access to. So it allows ...	[ { "start": 0, "end": 5.96, "text": " Hi there, today we'll look at inconsistency in conference peer review, revisiting the" }, { "start": 5.96, "end": 12.8, "text": " 2014 NeurIPS experiment by Corina Cortes and Neil D. Lawrence, which were actually the" }, { "start": 12.8, "...
W-O7AZNzbzQ	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	DDPM - Diffusion Models Beat GANs on Image Synthesis (Machine Learning Research Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "diffusion models", "diffusion model", "ddpm", "ddim", "denoising autoencoders", "generative models", "generative models deep learning", "gan alternatives", "altern...	#ddpm #diffusionmodels #openai GANs have dominated the image generation space for the majority of the last decade. This paper shows for the first time, how a non-GAN model, a DDPM, can be improved to overtake GANs at standard evaluation metrics for image generation. The produced samples look amazing and other than GAN...	Hello! These are generated images from a new model, actually a new class of model. It's been around for a while, but for the first time this new class of model has been pushed to the point where the images they produce not only look really nice and look like something you've come to expect from the latest and greatest...	[ { "start": 0, "end": 8.040000000000001, "text": " Hello! These are generated images from a new model, actually a new class of model." }, { "start": 8.040000000000001, "end": 13.44, "text": " It's been around for a while, but for the first time this new class of model has" }, { "s...
VQoyypYTz2U	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	All about AI Accelerators: GPU, TPU, Dataflow, Near-Memory, Optical, Neuromorphic & more (w/ Author)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "gpu", "tpu", "ipu", "wave computing", "dataflow", "near memory compute", "ai accelerators", "deep learning hardware", "sambanova", "cerebras", "graphcore", "...	#ai #gpu #tpu This video is an interview with Adi Fuchs, author of a series called "AI Accelerators", and an expert in modern AI acceleration technology. Accelerators like GPUs and TPUs are an integral part of today's AI landscape. Deep Neural Network training can be sped up by orders of magnitudes by making good use ...	Hello there! Today I'm talking to Adi Fuchs, who is an expert in AI acceleration technology. We talk about a whole bunch of things in this interview, but it is a little bit of a special thing because it's not about a paper or anything, but it is about a series of blog posts that Adi has authored. I am very much a noob...	[ { "start": 0, "end": 5.84, "text": " Hello there! Today I'm talking to Adi Fuchs, who is an expert in AI acceleration technology." }, { "start": 6.5600000000000005, "end": 11.52, "text": " We talk about a whole bunch of things in this interview, but it is a little bit of a special" }, ...
EA96xh9qog0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	I'm at ICML19 :)	[ "Science & Technology" ]	[ "machine learning", "conference", "long beach", "california", "icml19", "icml", "artificial intelligence", "ai", "deep learning" ]	Short intro to the International Conference on Machine Learning in Long Beach, CA. I'll be making some updates from the conference.	Hi there, it's day one of ICML and we'll be attending the conference here and just quickly pre-video to let everyone know I'll be trying to report from here kind of what papers are cool, what I liked, what are kind of the trends and so hopefully get this conference out to a broader community. So everyone's conglomerat...	[ { "start": 0, "end": 12.4, "text": " Hi there, it's day one of ICML and we'll be attending the conference here and just" }, { "start": 12.4, "end": 19.28, "text": " quickly pre-video to let everyone know I'll be trying to report from here kind of what" }, { "start": 19.28, "e...
-MCYbmU9kfg	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	RoBERTa: A Robustly Optimized BERT Pretraining Approach	[ "Science & Technology" ]	[ "deep learning", "machine learning", "nlp", "natural language processing", "machine translation", "arxiv", "google", "attention mechanism", "attention", "transformer", "tensor2tensor", "rnn", "recurrent", "seq2seq", "bert", "unsupervised", "squad", "wordpiece", "embeddings", "l...	This paper shows that the original BERT model, if trained correctly, can outperform all of the improvements that have been proposed lately, raising questions about the necessity and reasoning behind these. Abstract: Language model pretraining has led to significant performance gains but careful comparison between diff...	Hello everyone, today we're looking at Roberta, a robustly optimized BERT pre-training approach by Yin-Han Liu at AL, mainly of Facebook research. So this paper is a pretty short, pretty simple paper and the main premise is we've seen a number of improvements over the initial BERT paper where different pre-training of...	[ { "start": 0, "end": 6.84, "text": " Hello everyone, today we're looking at Roberta, a robustly optimized BERT pre-training approach" }, { "start": 6.84, "end": 11.96, "text": " by Yin-Han Liu at AL, mainly of Facebook research." }, { "start": 11.96, "end": 18.84, "text":...
pPBqM4CKjUU	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Discriminating Systems - Gender, Race, and Power in AI	[ "Science & Technology" ]	[ "ai", "machine learning", "bias", "fairness", "ml fairness", "algorithmic bias", "algorithmic discrimination", "ai and society", "ainow", "google", "microsoft", "race", "gender", "stem", "pipeline", "gender gap", "diversity", "inclusion", "equity", "power" ]	TL;DR: - There exists both an unequal representation of people in the AI workforce as well as examples of societal bias in AI systems. - The authors claim that the former causally leads to the latter and vice versa. - To me, the report does not manage to make a strong enough argument for that claim. - I find the statem...	Hi there, today we're looking at discriminating systems, gender, race and power in AI by Sarah Myers-West, Meredith Whitaker and Kate Crawford of the AI Now Institute, which is a part of New York University or associated with it. This is not as much a paper as it is a report, kind of summarizing current literature and...	[ { "start": 0, "end": 7.5200000000000005, "text": " Hi there, today we're looking at discriminating systems, gender, race and power in AI by Sarah" }, { "start": 7.5200000000000005, "end": 14.72, "text": " Myers-West, Meredith Whitaker and Kate Crawford of the AI Now Institute, which is a...
PDRtyrVskMU	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Chip Placement with Deep Reinforcement Learning (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "reinforcement learning", "deep reinforcement learning", "gans", "gan", "deconvolution", "computer chip", "gpu", "tpu", "fpga", "netlist", "constrained", "goo...	The AI Singularity is here! Computers designing new computers! It takes human experts multiple weeks to design new computer chips. What looks like a large game of Tetris is actually a very complex optimization problem. This paper uses Deep Reinforcement Learning to solve this optimization both faster and better than hu...	Hi there! Today we're looking at Chip Placement with Deep Reinforcement Learning by Azalia Miroszajny, Anna Goldi and a long list of authors that I have no stamina to read down. I'm sorry. So this work is a cool application of reinforcement learning to the real world. And we're gonna go through it and the cool thing a...	[ { "start": 0, "end": 5.48, "text": " Hi there! Today we're looking at Chip Placement with Deep Reinforcement Learning" }, { "start": 5.48, "end": 11.88, "text": " by Azalia Miroszajny, Anna Goldi and a long list of authors that I have no" }, { "start": 11.88, "end": 18.96, ...
B9PL__gVxLI	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	DeepMind's AlphaFold 2 Explained! AI Breakthrough in Protein Folding! What we know (& what we don't)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "google", "deepmind", "deep mind", "alphago", "alphazero", "alphafold", "protein", "dna", "rna", "folding", "casp", "casp14", "alphafold 2", "blog", "ha...	#deepmind #biology #ai This is Biology's AlexNet moment! DeepMind solves a 50-year old problem in Protein Folding Prediction. AlphaFold 2 improves over DeepMind's 2018 AlphaFold system with a new architecture and massively outperforms all competition. In this Video, we take a look at how AlphaFold 1 works and what we ...	It will change everything. DeepMind solves 50 year old grand challenge. The game has changed. DeepMind's latest AI breakthrough achieves historic new milestone, helps solve how diseases invade cells, improve protein folding prediction, AI breakthrough it also wipes your butt automatically. It is the newest DeepMind bi...	[ { "start": 0, "end": 11, "text": " It will change everything. DeepMind solves 50 year old grand challenge. The game has changed." }, { "start": 11, "end": 21, "text": " DeepMind's latest AI breakthrough achieves historic new milestone, helps solve how diseases invade cells," }, { ...
kOy49NqZeqI	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures	[ "Science & Technology" ]	[ "machine learning", "ml", "ai", "artificial intellgence", "deepmind", "reinforcement learning", "deep rl", "a2c", "a3c", "actor", "critic", "distributed", "scale", "bias", "off-policy", "policy gradient", "deepmind lab", "vtrace" ]	Policy Gradient RL on a massively distributed scale with theoretical guarantees! Abstract: In this work we aim to solve a large collection of tasks using a single reinforcement learning agent with a single set of parameters. A key challenge is to handle the increased amount of data and extended training time. We have ...	Hi there! Today we're looking at Impala, scalable distributed deep RL with importance-weighted actor learner architectures by Lasse Espejolt, Hubert Sawyer, Remy Munoz and Al. So this paper deals with a new architecture for deep reinforcement learning, specifically distributed deep reinforcement learning. So that mean...	[ { "start": 0, "end": 5.88, "text": " Hi there! Today we're looking at Impala, scalable distributed deep RL with" }, { "start": 5.88, "end": 11.48, "text": " importance-weighted actor learner architectures by Lasse Espejolt, Hubert" }, { "start": 11.48, "end": 18.48, "text...
RrvC8YW0pT0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Reinforcement Learning Upside Down: Don't Predict Rewards -- Just Map Them to Actions	[ "Science & Technology" ]	[ "rl", "reinforcement learning", "ai", "artificial intelligence", "udrl", "schmidhuber", "policy", "value", "reward" ]	Schmidhuber thinking outside the box! Upside-Down RL turns RL on its head and constructs a behavior function that uses the desired reward as an input. The new paradigm shows surprising performance compared to classic RL algorithms. Abstract: We transform reinforcement learning (RL) into a form of supervised learning (...	He did it! Crazy son of a bitch did it again! What am I talking about? Jürgen Schmidhuber reinforcement learning upside down! New paper just dropped on the verge of the NeurIPS conference being presented at a workshop here. Presenting upside down reinforcement learning. I am pumped for this one, can you tell? It says ...	[ { "start": 0, "end": 6.4, "text": " He did it! Crazy son of a bitch did it again!" }, { "start": 6.4, "end": 12.8, "text": " What am I talking about? Jürgen Schmidhuber reinforcement learning upside down!" }, { "start": 12.8, "end": 20.6, "text": " New paper just dropped ...
hsOMCwvFv80	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	I'm out of Academia	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper" ]	#machinelearning #ai #phd Done with my PhD in Machine Learning at ETH Zurich. On to new lands! Links: TabNine Code Completion (Referral): http://bit.ly/tabnine-yannick YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://discord.gg/4H8xxDF BitChute: https://www.bitch...	Howdy diddly doo. Hi everyone. If you're wondering what the ridiculous thing on my head is, then that is my official graduation slash successful defense hat. I'm not yet allowed to technically use the title Doctor but let's be honest who gives a crap anyway about titles. I'm a huge fan of this hat my lab mates made th...	[ { "start": 0, "end": 6.32, "text": " Howdy diddly doo. Hi everyone. If you're wondering what the ridiculous thing on my head is," }, { "start": 6.88, "end": 15.76, "text": " then that is my official graduation slash successful defense hat. I'm not yet allowed to" }, { "start": 15...
X4S8F3bwuuw	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Author Interview: SayCan - Do As I Can, Not As I Say: Grounding Language in Robotic Affordances	[ "Science & Technology" ]	[]	#saycan #robots #ai This is an interview with the authors Brian Ichter, Karol Hausman, and Fei Xia. Original Paper Review Video: https://youtu.be/Ru23eWAQ6_E Large Language Models are excellent at generating plausible plans in response to real-world problems, but without interacting with the environment, they have no ...	So today we're here with three of the authors of this paper with I have to say a lot of authors It seems like a giant work just from what I could gather From the from the paper itself and the data collection and the evaluation and so on So this was a huge thing, but the results are pretty cool. So here with me today a...	[ { "start": 0, "end": 6.140000000000001, "text": " So today we're here with three of the authors of this paper with I have to say a lot of authors" }, { "start": 6.3, "end": 9.06, "text": " It seems like a giant work just from what I could gather" }, { "start": 9.540000000000001, ...
qgUegkefocg	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Fastformer: Additive Attention Can Be All You Need (Machine Learning Research Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "attention mechanism", "attention is all you need", "fastformer", "fast former", "nlp", "natural language processing", "linear attention", "linear transformer", "qu...	#attention #transformer #fastformer Transformers have become the dominant model class in the last few years for large data, but their quadratic complexity in terms of sequence length has plagued them until now. Fastformer claims to be the fastest and most performant linear attention variant, able to consume long conte...	Hello there! Today we'll look at Fastformer Additive Attention Can Be All You Need by Chuan Wu, Fang Zhao Wu, Tao Qi, and Yongfeng Huang. So this paper definitely wins out in the category of most innovative paper titles of the last few months, as apparently we've gone from Is All You Need to Can Be All You Need. So a ...	[ { "start": 0, "end": 6.16, "text": " Hello there! Today we'll look at Fastformer Additive Attention Can Be All You Need by" }, { "start": 6.16, "end": 14.120000000000001, "text": " Chuan Wu, Fang Zhao Wu, Tao Qi, and Yongfeng Huang. So this paper definitely wins out in the category" },...
_c6A33Fg5Ns	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	DeBERTa: Decoding-enhanced BERT with Disentangled Attention (Machine Learning Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "deep learning tutorial", "huggingface", "huggingface transformers", "microsoft", "microsoft research", "bert", "roberta", "deberta", "nlp", "natural language pro...	#deberta #bert #huggingface DeBERTa by Microsoft is the next iteration of BERT-style Self-Attention Transformer models, surpassing RoBERTa in State-of-the-art in multiple NLP tasks. DeBERTa brings two key improvements: First, they treat content and position information separately in a new form of disentangled attentio...	Hi there, today we'll look at DeBURTA, decoding enhanced BERT with disentangled attention, by Peng Cheng He, Xia Dong Liu, Zhang Feng Gao, and Wai Ju Chen of Microsoft. This paper is an improvement on BERT, the language model and the Roburta variant of it. Specifically, it suggests two improvements, namely, first is t...	[ { "start": 0, "end": 7.5200000000000005, "text": " Hi there, today we'll look at DeBURTA, decoding enhanced BERT with disentangled attention," }, { "start": 7.5200000000000005, "end": 14.36, "text": " by Peng Cheng He, Xia Dong Liu, Zhang Feng Gao, and Wai Ju Chen of Microsoft." }, {...
69IjNZaoeao	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	LeDeepChef 👨‍🍳 Deep Reinforcement Learning Agent for Families of Text-Based Games	[ "Science & Technology" ]	[ "ml", "machine learning", "reinforcement learning", "recipe", "text-based games", "text games", "natural language processing", "nlp", "actor", "critic", "GRU", "embedding", "pretraining", "artificial intelligence", "ai", "competition", "microsoft" ]	The AI cook is here! This agent learns to play a text-based game where the goal is to prepare a meal according to a recipe. Challenges? Many! The number of possible actions is huge, ingredients change and can include ones never seen before, you need to navigate rooms, use tools, manage an inventory and sequence everyth...	Hi there. Today we're looking at Le Deep Chef, deep reinforcement learning agent for families of text-based games by Leonard Adolfs and Thomas Hoffmann. So this is a paper about engineering an agent for a particular family of tasks. This is different from reinforcement learning agents that for example are just good at...	[ { "start": 0, "end": 5.4, "text": " Hi there. Today we're looking at Le Deep Chef, deep reinforcement learning agent" }, { "start": 5.4, "end": 11.28, "text": " for families of text-based games by Leonard Adolfs and Thomas Hoffmann. So" }, { "start": 11.28, "end": 18.40000000...
_PyusGsbBPY	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Stochastic RNNs without Teacher-Forcing	[ "Science & Technology" ]	[ "NeurIPS2018", "NIPS2018", "NLP", "deep learning", "RNN" ]	We present a stochastic non-autoregressive RNN that does not require teacher-forcing for training. The content is based on our 2018 NeurIPS paper: Deep State Space Models for Unconditional Word Generation https://arxiv.org/abs/1806.04550	Hi everybody, my name is Florian and Janik was nice enough to host me here as a guest to talk about Stochastic RNNs without teacher forcing. This is based on recent work, deep state space models for unconditional word generation, which we presented at this year's New RIPs. And if you feel like any more details, please...	[ { "start": 0, "end": 6, "text": " Hi everybody, my name is Florian and Janik was nice enough to host me here as a guest to talk about" }, { "start": 6, "end": 14, "text": " Stochastic RNNs without teacher forcing. This is based on recent work, deep state space models for" }, { "s...
-YiMVR3HEuY	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Reinforcement Learning with Unsupervised Auxiliary Tasks	[ "Science & Technology" ]	[ "machine learning", "artificial intelligence", "ai", "deep learning", "unsupervised learning", "research", "academia", "paper", "review", "agents", "tasks" ]	ERROR: type should be string, got "https://arxiv.org/abs/1611.05397\n\nAbstract:\nDeep reinforcement learning agents have achieved state-of-the-art results by directly maximising cumulative reward. However, environments contain a much wider variety of possible training signals. In this paper, we introduce an agent that also maximises many other pseudo-reward functions simultaneously by reinforcement learning. All of these tasks share a common representation that, like unsupervised learning, continues to develop in the absence of extrinsic rewards. We also introduce a novel mechanism for focusing this representation upon extrinsic rewards, so that learning can rapidly adapt to the most relevant aspects of the actual task. Our agent significantly outperforms the previous state-of-the-art on Atari, averaging 880\\% expert human performance, and a challenging suite of first-person, three-dimensional \\emph{Labyrinth} tasks leading to a mean speedup in learning of 10× and averaging 87\\% expert human performance on Labyrinth.\n\nAuthors:\nMax Jaderberg, Volodymyr Mnih, Wojciech Marian Czarnecki, Tom Schaul, Joel Z Leibo, David Silver, Koray Kavukcuoglu"	Hi there, today we're looking at reinforcement learning with unsupervised auxiliary tasks by Google. So in this paper the authors consider a reinforcement learning task and I can show you what it looks like. It looks like this kind of a maze or this is an example that they give where you have to navigate the maze, it'...	[ { "start": 0, "end": 6.48, "text": " Hi there, today we're looking at reinforcement learning with unsupervised auxiliary tasks" }, { "start": 6.48, "end": 9.64, "text": " by Google." }, { "start": 9.64, "end": 14.6, "text": " So in this paper the authors consider a reinfo...
xTzFJIknh7E	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	TransCoder: Unsupervised Translation of Programming Languages (Paper Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper" ]	Code migration between languages is an expensive and laborious task. To translate from one language to the other, one needs to be an expert at both. Current automatic tools often produce illegible and complicated code. This paper applies unsupervised neural machine translation to source code of Python, C++, and Java an...	Hi there! So the paper we're looking at today can take the code on the left, which is written in Python, and can output the code on the right, which is written in C++. Now the point here is that the code on the right does the same thing as the code on the left, so it is implementing the same function. The surprising t...	[ { "start": 0, "end": 5.24, "text": " Hi there! So the paper we're looking at today can take the code on the left," }, { "start": 5.24, "end": 9.8, "text": " which is written in Python, and can output the code on the right, which is" }, { "start": 9.8, "end": 15.16, "text"...
pZyxlf6l0N8	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	Thinking While Moving: Deep Reinforcement Learning with Concurrent Control	[ "Science & Technology" ]	[ "deep learning", "machine learning", "reinforcement learning", "vector to go", "vtg", "continuous", "control", "robot", "concurrent", "deep rl", "deep neural networks", "berkeley", "google", "grasping", "qlearning" ]	Classic RL "stops" the world whenever the Agent computes a new action. This paper considers a more realistic scenario where the agent is thinking about the next action to take while still performing the last action. This results in a fascinating way of reformulating Q-learning in continuous time, then introducing concu...	Hi there. So if you look at these two robots, the left one labeled blocking, the right one labeled concurrent, the blocking robot, as you can see, always has these little pauses in its movement where it does nothing and then it kind of continues with its motion, while the one on the right is one continuous motion that...	[ { "start": 0, "end": 7, "text": " Hi there. So if you look at these two robots, the left one labeled blocking, the right one labeled concurrent," }, { "start": 7, "end": 15, "text": " the blocking robot, as you can see, always has these little pauses in its movement where it does nothing...
5skIqoO3ku0	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	OpenAI Embeddings (and Controversy?!)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "natural language processing", "mlnews", "openai", "openai embeddings", "nils reimers", "beir dataset", "beir benchmark", "text similarity", "neural embeddings", ...	#mlnews #openai #embeddings COMMENTS DIRECTLY FROM THE AUTHOR (thanks a lot for reaching out Arvind :) ): 1. The FIQA results you share also have code to reproduce the results in the paper using the API: https://twitter.com/arvind_io/status/1488257004783112192?s=20&t=gB3c79VEX8hGJl6WfZa2iA There's no discrepancy AFAIK...	Hello, everyone, welcome to a special edition of ML news, we have something to discuss. Open AI just released an embeddings endpoint to their API. This is a company by blog post called introducing text and code embeddings in the Open AI API. Now after the let's call them big successes of GPT three, and codecs, which i...	[ { "start": 0, "end": 11, "text": " Hello, everyone, welcome to a special edition of ML news, we have something to discuss." }, { "start": 11, "end": 15.24, "text": " Open AI just released an embeddings endpoint to their API." }, { "start": 15.24, "end": 21.52, "text": " T...
j4xgkjWlfL4	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	OpenAI DALL·E: Creating Images from Text (Blog Post Explained)	[ "Science & Technology" ]	[ "deep learning", "machine learning", "arxiv", "explained", "neural networks", "ai", "artificial intelligence", "paper", "gpt", "gpt-3", "visual transformer", "transformer", "transformers", "attention mechanism", "vqvae", "vq vae", "vq-vae", "codebook", "relaxation", "gumbel", ...	#openai #science #gpt3 OpenAI's newest model, DALL·E, shows absolutely amazing abilities in generating high-quality images from arbitrary text descriptions. Like GPT-3, the range of applications and the diversity of outputs is astonishing, given that this is a single model, trained on a purely autoregressive task. Thi...	A sphere made of Swiss cheese. A sphere with a texture of Swiss cheese. And there you have it. Beautiful, very appetizing Swiss cheese balls. My Swiss heart had just skipped a beat out of this monstrosity. What's even cooler than a sphere made of Swiss cheese is a torus made of denim. These images are so cool. A torus...	[ { "start": 0, "end": 9, "text": " A sphere made of Swiss cheese. A sphere with a texture of Swiss cheese." }, { "start": 9, "end": 17.76, "text": " And there you have it. Beautiful, very appetizing Swiss cheese balls. My Swiss heart had just" }, { "start": 17.76, "end": 25.04...
yPjuAo53uNI	Yannic Kilcher	UCZHmQk67mSJgfCCTn7xBfew	[Rant] The Male Only History of Deep Learning	[ "Science & Technology" ]	[ "deep learning", "machine learning", "neural networks", "history", "groups", "ideology" ]	This casting of our field in terms of ideological narrow-sighted group-think is disgusting. Keep Science about ideas! https://twitter.com/timnitGebru/status/1252752743942328321 Links: YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher BitChute: https://www.bitchute.com/channel/yann...	Alright, so instead of reviewing a paper today, I thought I might review this thing. So this person on Twitter posted this link to an article called Brief History of Deep Learning from 1943 to 2019 of Machine Learning Knowledge.ai. So let's look at this. Actually let's look at the tweet first. Because this is... I jus...	[ { "start": 0, "end": 6.26, "text": " Alright, so instead of reviewing a paper today, I thought I might review this thing." }, { "start": 6.26, "end": 13.22, "text": " So this person on Twitter posted this link to an article called Brief History of Deep" }, { "start": 13.22, "...

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.