Chaînes mises à jour récemment

Choisissez-en une pour ne voir qu’elleToutes les chaînes →

Résultats

14 articles
Trier
2:02:17EN
Cognitive Revolution "How AI Changes Everything"· 27 août 2025PodcastSAEs Find the Nouns. Parameter Decomposition Looks for the VerbsAnime : Nathan LabenzGoodfire's Lee Sharkey argues that sparse autoencoders recover a network's variables, not the algorithms those variables run, and shows how stochastic parameter decomposition tries to recover the computations themselves.Modèles et entraînementModel InterpretabilityAI research+3
1:37:05EN
Cognitive Revolution "How AI Changes Everything"· 19 juin 2025PodcastAn Algebraic Geometer Quit Tenure Because the Bowl-Shaped Loss Plot Is a LieAnime : Nathan LabenzJesse Hoogland and Daniel Murfet of Timaeus argue that singularities in the loss landscape — not the smooth valleys in every slide deck — are what actually govern how neural networks learn, generalize, and go wrong.Modèles et entraînementModel InterpretabilityFondamentaux du ML+3
2:12:00EN
Cognitive Revolution "How AI Changes Everything"· 15 mai 2025PodcastGradient Descent at Test Time: Titans Pushes LLM Context to 10 Million TokensAnime : Nathan LabenzA Cornell PhD student's neural memory module holds about 70% accuracy where GPT-4's scores fall off a cliff.Modèles et entraînementAgent memoryIngénierie du contexte+3
2:33:39EN
Cognitive Revolution "How AI Changes Everything"· 27 févr. 2025PodcastThe Attacker Goes Second: Why Nicholas Carlini Keeps Breaking AI DefensesAnime : Nathan LabenzA Google DeepMind security researcher on simple loss functions, 70% robust accuracy, and why the fix has to live in the system, not the model.SécuritéAI securitySécurité et gouvernance+3
1:40:22EN
Cognitive Revolution "How AI Changes Everything"· 22 janv. 2025PodcastTwenty-atom crystals taught an AI the physics of a heartbeatAnime : Nathan LabenzOrbital Materials' Jonathan Godwin and Tim Duignan tell Nathan Labenz how a neural network trained on tiny crystals simulated a potassium ion channel on a single 2017 GPU.Modèles et entraînementAI for scienceFondamentaux du ML+3
1:42:16EN
Cognitive Revolution "How AI Changes Everything"· 4 juil. 2024Podcast267 Papers in Seven Months: Albert Gu Explains the State Space Model RevolutionAnime : Nathan LabenzThe Mamba co-author walks through five years of recurrent-model work, why selective compression beats a growing KV cache, and what Mamba 2 trades away to train faster.Modèles et entraînementTransformers & attentionInference+3
1:54:01EN
Cognitive Revolution "How AI Changes Everything"· 26 juin 2024PodcastA Tiny Network Grew a Salt Crystal the Training Data Never ShowedAnime : Nathan LabenzTim Duignan trained tens of thousands of parameters on salt water. The model invented a crystal — and a continuum solvent theory he had spent a PhD failing to write by hand.Modèles et entraînementAI for scienceFondamentaux du ML+3
1:58:43EN
Cognitive Revolution "How AI Changes Everything"· 7 févr. 2024PodcastA Worm With 302 Neurons Already Makes Trade-offs. AI Is Still on the Same Ladder.Anime : Nathan LabenzMax Bennett maps 600 million years of brain evolution onto five breakthroughs — and asks which rungs today's models have actually reached.Modèles et entraînementAI HistoryNeuroscience & AI+3
1:19:24EN
Cognitive Revolution "How AI Changes Everything"· 17 nov. 2023PodcastA Toy Paper Did Not Prove LLMs Cannot Generalize. The Discourse Still Ran With It.Anime : Nathan LabenzNathan Labenz and Erik Torenberg unpack a viral DeepMind result, a weekend reproduction, and a premature war over voluntary AI commitments.Modèles et entraînementAI researchFondamentaux du ML+3
1:33:56EN
Cognitive Revolution "How AI Changes Everything"· 6 juil. 2023PodcastThe AI Scouting Report: 90 Minutes From GPT-4's Exam Scores to Whether Models UnderstandAnime : Nathan LabenzNathan Labenz spends a solo episode walking through what general-purpose models can do, how they are trained, how capabilities unlock, and whether they understand anything at all.Modèles et entraînementFondamentaux du MLScaling Laws+3
1:32:22EN
Cognitive Revolution "How AI Changes Everything"· 29 juin 2023PodcastA Million Bytes and 256 Symbols: Meta's MEGABYTE Wants to Throw the Tokenizer AwayAnime : Nathan LabenzLili Yu of Meta AI explains a multiscale transformer that predicts million-byte sequences, skips tokenization, and aims to model any file as raw bytes.Modèles et entraînementMultimodal AITransformers & attention+3
1:52:46EN
Cognitive Revolution "How AI Changes Everything"· 27 juin 2023PodcastCharge Neural Nets by Connection Length, and the Black Box Splits Into Modules You Can SeeAnime : Nathan LabenzMIT physicist Ziming Liu walks Nathan Labenz through BIMT, a brain-inspired training trick that sparsifies networks until you can read them by looking.Modèles et entraînementModel InterpretabilityFondamentaux du ML+3
1:25:50EN
Cognitive Revolution "How AI Changes Everything"· 18 avr. 2023PodcastPrompt Engineering Became Programming, and Riley Goodside Is Fine With ThatAnime : Nathan LabenzScale AI’s first staff prompt engineer on tool use, fine-tuning without amnesia, generate-then-rank, and why the 2023 job-of-the-future story never arrived.Entreprise et stratégiePrompting & in-context learningStructured outputs+3
1:03:10EN
Cognitive Revolution "How AI Changes Everything"PodcastPause Tokens, Analogical Prompts and Ring Attention: A Tour of the Transformer's Next UpgradesAnime : Nathan LabenzNathan Labenz and Trey Kollmer walk through five recent papers on pausing, backspacing, prompting, streaming and long-context training, and sketch how they might combine into a more capable model.Modèles et entraînementTransformers & attentionFondamentaux du ML+3