Yannic Kilcher Videos (Audio Only)

#deeplearning #co2 #cost Deep Learning has achieved impressive results in the last years, not least due to the massive increases in computational power and data that has gone into these models. Scaling up currently promises to be a reliable way to create more performant systems, but how far can we go? This article explores the limits of exponential scaling in AI, and what people are doing to get a...

#plagiarism #surveillance #schmidhuber Your Mondaily updates of what's going in the world of Machine Learning. OUTLINE: 0:00 - Intro 0:20 - New plagiarism case has plot twist 7:25 - CLIP for video surveillance 9:40 - DARPA SubTerranean Challenge 11:00 - Schmidhuber criticizing Turing Lecture 15:00 - OpenAI summarizes books 17:55 - UnBiasIt monitors employees' communications for bias 20:00 - iOS pl...

#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 accepted papers were consistently accepted by both committees and demonstrated signif...

#truthfulqa #efficientnet #laion400M Your regularly irregular updates on what's happening in the Machine Learning world. OUTLINE: 0:00 - Intro 0:20 - TruthfulQA benchmark shines new light on GPT-3 2:00 - LAION-400M image-text-pair dataset 4:10 - GoogleAI's EfficientNetV2 and CoAtNet 6:15 - Uber's H3: A hexagonal coordinate system 7:40 - AWS NeurIPS 2021 DeepRacer Challenge 8:15 - Helpful Libraries...

#gpt-3 #truth #conspiracy A new benchmark paper has created quite an uproar in the community. TruthfulQA is a dataset of 817 questions probing for imitative falsehoods where language models become less truthful, the larger they get. This surprising counter-intuitive finding validates many people's criticisms of large language models, but is it really the correct conclusion? OUTLINE: 0:00 - Intro 0...

#tvae #topographic #equivariant Variational Autoencoders model the latent space as a set of independent Gaussian random variables, which the decoder maps to a data distribution. However, this independence is not always desired, for example when dealing with video sequences, we know that successive frames are heavily correlated. Thus, any latent space dealing with such data should reflect this in i...

#schmidhuber #tiktok #roomba Your regularly irregular update on what's happening in the world of Machine Learning. OUTLINE: 0:00 - Intro 0:15 - Sponsor: Weights & Biases 1:55 - ML YouTuber reaches 100k subscribers 2:40 - Facebook AI pushes Textless NLP 5:30 - Schmidhuber blog post: I invented everything 7:55 - TikTok algorithm rabbitholes users 10:45 - Roomba learns to avoid poop 11:50 - AI ca...

#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 BitChute: https://www.bitchute.com/channel/yann... Minds: https://www.minds.com/ykil...

#mlnews #schmidhuber #muzero Your regular updates on what's happening in the ML world! OUTLINE: 0:00 - Intro 0:15 - Sponsor: Weights & Biases 1:45 - Google shuts down health streams 4:25 - AI predicts race from blurry X-Rays 7:35 - Facebook labels black men as primates 11:05 - Distill papers on Graph Neural Networks 11:50 - Jürgen Schmidhuber to lead KAUST AI Initiative 12:35 - GitHub brief on...

#inftyformer #infinityformer #transformer Vanilla Transformers are excellent sequence models, but suffer from very harsch constraints on the length of the sequences they can process. Several attempts have been made to extend the Transformer's sequence length, but few have successfully gone beyond a constant factor improvement. This paper presents a method, based on continuous attention mechanisms,...

#mlnews #chess #neurips OUTLINE: 0:00 - Intro 0:30 - Reconnaissance Blind Chess NeurIPS 2021 Competition 3:40 - Colab Pro no longer top priority for GPUs 4:45 - DeepMind uses Graph NNs to do traffic prediction 6:00 - Helpful Libraries: Isaac Gym, Differentiable Human, LVIS, BEHAVIOR 10:25 - Cerebras Wafer Scale Engine Cluster 12:15 - AI Voice Synthesis for Val Kilmer 14:20 - Can AI give thoughtful...

#alibi #transformers #attention Transformers are essentially set models that need additional inputs to make sense of sequence data. The most widespread additional inputs are position encodings or position embeddings, which add sequence index information in various forms. However, this has put a limit on the resulting model, which cannot run inference on sequences longer than it has been trained on...

#plagiarism #foundationmodels #tesla The best place to keep up to date with the latest and greatest from the ML world! OUTLINE: 0:00 - Intro & Sponsor 3:15 - A high-profile case of plagiarism shocks the ML world 11:55 - Stanford AI releases paper on "Foundation Models" 19:45 - Updates on Apple's NeuralHash 20:45 - RL control for two-player splorts 21:45 - Tesla's AI Day 23:55 - COMMA THREE ann...

#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 contexts at once. This is achieved by a combination of additive attention and elementw...

#pondernet #deepmind #machinelearning Humans don't spend the same amount of mental effort on all problems equally. Instead, we respond quickly to easy tasks, and we take our time to deliberate hard tasks. DeepMind's PonderNet attempts to achieve the same by dynamically deciding how many computation steps to allocate to any single input sample. This is done via a recurrent architecture and a traina...

#apple #icloud #neuralhash Send your Apple fanboy friends to prison with this one simple trick ;) We break Apple's NeuralHash algorithm used to detect CSAM for iCloud photos. I show how it's possible to craft arbitrary hash collisions from any source / target image pair using an adversarial example attack. This can be used for many purposes, such as evading detection, or forging false positives, t...

#mlnews #nvidia #openai An in-depth look over what's going on in the world of Machine Learning and Artificial intelligence. Subscribe now and make Monday the best day of the week! OUTLINE: 0:00 - Intro 0:20 - Sponsor: Weights & Biases 3:00 - Nvidia's CEO was rendered during Keynote 5:00 - AI21 Labs releases Jurassic-1 language model 7:00 - Tortured Phrases reveal plagiarism 10:05 - Cortical ne...

#apple #icloud #privacy Apple recently announced scanning all images uploaded to iCloud for CSAM (child abuse material), and that this scan would happen locally on users' phones. We take a look at the technical report and explore how the system works in detail, how it is designed to preserve user privacy, and what weak points it still has. OUTLINE: 0:00 - Introduction 3:05 - System Requirements 9:...

#mlnews #apple #nolamarck Your update on the latest news in the AI and Machine Learning world. OUTLINE: 0:00 - Intro 0:15 - Sponsor: Weights & Biases 3:30 - Apple to scan iDevices for illegal content 14:10 - EU approves chatcontrol 15:20 - Machine Learning FAQ book 17:40 - TimeDial & Disfl-QA Conversation Datasets 20:30 - VoxPopuli Speech Dataset 21:00 - Google Tensor chip coming to Pixel...

#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 app 16:15 - Wordcraft interactive AI text editor 17:05 - ML used to cheat in...

#chai #mlnews #nvidia Follow Saynam here: YouTube: https://www.youtube.com/c/ChaiTimeDat... Twitter: https://twitter.com/bhutanisanyam1 Apple Podcasts: https://podcasts.apple.com/us/podcast... LinkedIn: https://www.linkedin.com/in/sanyambhu... Spotify: https://open.spotify.com/show/7IbEWJj... Anchor.fm RSS: https://anchor.fm/s/c19772c/podcast/rss Outline: 0:00 - Intro & Overview 1:30 - Amazon'...

A look into the happenings of the Machine Learning world. OUTLINE: 0:00 - Intro 0:25 - Facebook AI trains rapidly adapting robots 3:05 - Baidu presents autonomous excavator system 4:45 - EleutherAI turns 1 6:05 - Elon Musk says FSD harder than expected 8:10 - AI interview tools still fall short 11:10 - RunwayML AI-powered cloud video editor 11:55 - MineRL BASALT competition to learn from human fee...

#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 licenses, social obligations, and how much you can profit from other people's wor...

#tesla #selfdriving #karpathy Tesla is pushing the state-of-the-art in full self-driving, and interestingly, they explicitly switch from having multiple different sensors to a vision-only system. We discuss the highlights of Andrej Karpathy's talk about Tesla's FSD system, how to label petabytes of data, how to sample edge-cases, how to train a neural network that has to work in real-time, and why...

#cvpr #socialmedia #machinelearning In this week's ML news we look at CVPR's controversial action to ban paper promotions on social media during the review phase, among other things! OUTLINE: 0:00 - Intro & Overview 0:25 - CVPR bans social media paper discussions 5:10 - WalMart uses AI to suggest substitutions 6:05 - NVIDIA releases Alias-Free GAN 7:30 - Confession Video in Myanmar possibly a...