Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

I had an awesome time interviewing idan shenfeld and Jonas Hübotter from MIT and ETH Zurich about self-distillation. this very promising post-training paradigm where the model acts as its own teacher by conditioning on environment feedback or demonstrations. we cover the SDPO algo for reinforcement learning with rich feedback... and SDFT for continual learning without forgetting along with many applications. we dig into how it works, why it's simpler and faster than GRPO, and where this is already showing up in production systems. table of content: 0:00 - what is self distillation 2:50 - idan (MIT) and jonas (ETH Zurich) introduction and motivation 18:40 - different perspective of on-policy self-distillation (presentation) 36:00 - metacognition and specificity in self-distillation 37:24 - very long hard task and self-distillation 42:00 - continual learning with self-distillation (presentation) 1:16:50 - what is next in this research direction? 1:20:00 - is there any experience with subjective feedbacks? 1:22:50 - quality vs number of feedbacks? 1:26:40 - what setting would self-distillation struggle vs GRPO? the slides were super crisp really cool of them to share! enjoy my guys 🌹show more

Yacine Mahdid

29,769 subscribers

13,091 Aufrufe • vor 4 Monaten •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

New lecture for the book! Nominally about synthetic data, but mostly is a walk through of the distillation literature from the Hinton 2015 paper to multi-teach on-policy distillation of today! At 7.4 hours of video in my post-training brain dump and counting :) It was fun to stare at the math long enough and talk through the 3-4 core changes that needed to be made to the original formulation to have on-policy distillation be ready for the mainstream like it is today (and in RL frameworks). Otherwise, I include a bit of a history lesson for how synthetic data generally slowly took over all post-training data research (it wasn't always the case)! Then I do some 101 review on constitutional AI, rubrics, and other popular methods. 00:00 The emergence of synthetic data 10:50 Background on teacher-student knowledge-distillation 24:47: On-policy distillation (OPD, MOPD, and OPSD) 37:11 Constitutional AI & AI Feedback 45:50 Rubrics as rewards & conclusions Ofc, watch on YouTube etc.

New lecture for the book! Nominally about synthetic data, but mostly is a walk through of the distillation literature from the Hinton 2015 paper to multi-teach on-policy distillation of today! At 7.4 hours of video in my post-training brain dump and counting :) It was fun to stare at the math long enough and talk through the 3-4 core changes that needed to be made to the original formulation to have on-policy distillation be ready for the mainstream like it is today (and in RL frameworks). Otherwise, I include a bit of a history lesson for how synthetic data generally slowly took over all post-training data research (it wasn't always the case)! Then I do some 101 review on constitutional AI, rubrics, and other popular methods. 00:00 The emergence of synthetic data 10:50 Background on teacher-student knowledge-distillation 24:47: On-policy distillation (OPD, MOPD, and OPSD) 37:11 Constitutional AI & AI Feedback 45:50 Rubrics as rewards & conclusions Ofc, watch on YouTube etc.

Nathan Lambert

93,111 Aufrufe • vor 2 Monaten

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!

Jürgen Schmidhuber

249,595 Aufrufe • vor 6 Monaten

Here's my conversation with Charan Ranganath, psychologist & neuroscientist at UC Davis, specializing in human memory. We talk about memory, imagination, deja vu, false memories, why we remember, why we forget, and a lot of other fascinating mysteries about the human mind. It's here on X in full, and is up on YouTube, Spotify, and everywhere else. Links in comment. Timestamps: 0:00 - Introduction 1:03 - Experiencing self vs remembering self 14:44 - Creating memories 24:16 - Why we forget 31:53 - Training memory 42:22 - Memory hacks 54:10 - Imagination vs memory 1:03:29 - Memory competitions 1:13:18 - Science of memory 1:28:33 - Discoveries 1:39:37 - Deja vu 1:44:54 - False memories 2:04:59 - False confessions 2:08:45 - Heartbreak 2:16:19 - Nature of time 2:24:00 - Brain–computer interface (BCI) 2:38:04 - AI and memory 2:48:18 - ADHD 2:55:15 - Music 3:05:00 - Human mind

Here's my conversation with Charan Ranganath, psychologist & neuroscientist at UC Davis, specializing in human memory. We talk about memory, imagination, deja vu, false memories, why we remember, why we forget, and a lot of other fascinating mysteries about the human mind. It's here on X in full, and is up on YouTube, Spotify, and everywhere else. Links in comment. Timestamps: 0:00 - Introduction 1:03 - Experiencing self vs remembering self 14:44 - Creating memories 24:16 - Why we forget 31:53 - Training memory 42:22 - Memory hacks 54:10 - Imagination vs memory 1:03:29 - Memory competitions 1:13:18 - Science of memory 1:28:33 - Discoveries 1:39:37 - Deja vu 1:44:54 - False memories 2:04:59 - False confessions 2:08:45 - Heartbreak 2:16:19 - Nature of time 2:24:00 - Brain–computer interface (BCI) 2:38:04 - AI and memory 2:48:18 - ADHD 2:55:15 - Music 3:05:00 - Human mind

Lex Fridman

294,716 Aufrufe • vor 2 Jahren

The full Transformer vs Post-Transformer debate is live. 80 minutes. Seven rounds. No slides. Real disagreement. Lukasz Kaiser came to defend the Transformer. Adrian Kosowski, Llion Jones, and Mathias Lechner made the case for what comes next. 00:00 Contenders enter the ring 06:30 Lukasz Kaiser defends the Transformer 10:08 Adrian Kosowski on BDH and the PageRank Moment for AI 17:35 Llion Jones: Why Transformers aren't the final architecture 29:50 Mathias Lechner on Liquid AI’s approach, Fast Weights, and Self-Replacing AI 40:28 Reasoning Beyond Language 44:15 Scaling Laws: Transformer vs Post Transformer 50:31 Benchmarks, Coding Models, and Perplexity 1:04:00 Continual Learning and Dynamic Weights This is the ultimate source of truth on the subject.

The full Transformer vs Post-Transformer debate is live. 80 minutes. Seven rounds. No slides. Real disagreement. Lukasz Kaiser came to defend the Transformer. Adrian Kosowski, Llion Jones, and Mathias Lechner made the case for what comes next. 00:00 Contenders enter the ring 06:30 Lukasz Kaiser defends the Transformer 10:08 Adrian Kosowski on BDH and the PageRank Moment for AI 17:35 Llion Jones: Why Transformers aren't the final architecture 29:50 Mathias Lechner on Liquid AI’s approach, Fast Weights, and Self-Replacing AI 40:28 Reasoning Beyond Language 44:15 Scaling Laws: Transformer vs Post Transformer 50:31 Benchmarks, Coding Models, and Perplexity 1:04:00 Continual Learning and Dynamic Weights This is the ultimate source of truth on the subject.

Zuzanna Stamirowska

1,378,255 Aufrufe • vor 4 Monaten

Are Chinese open models being oversold as reaching the frontier? We’re back with another vibe check episode with Ari Morcos and Rob Toews. These are a ton of fun to record and this was a particularly meaty one given everything going on. We hit on: ▪️Did Chinese open models catch the US frontier? ▪️How much does distillation explain China's progress ▪️Reactions to Anthropic Backlash ▪️What a Stuxnet style backdoor in model weights looks like ▪️The OpenAI and Hugging Face breach ▪️xAI and Cursor ▪️Prediction check-ins 0:00 Intro 2:41 China's Open Source Models Catch Up 7:42 Does Distillation Explain China's Rise? 13:54 The Geopolitical Risk of Chinese AI Models 20:28 Should the Government Restrict Open Models? 22:06 What Are the Labs Really Learning From You? 29:22 Future of Government Regulation 40:50 The OpenAI-Hugging Face Hack 47:59 Grok, Cursor, and the Value of Real Data 56:52 SSI and OpenRouter 1:02:52 Venture Capital's Return to Deep Tech 1:05:16 Quickfire YouTube: Spotify: Apple:

Are Chinese open models being oversold as reaching the frontier? We’re back with another vibe check episode with Ari Morcos and Rob Toews. These are a ton of fun to record and this was a particularly meaty one given everything going on. We hit on: ▪️Did Chinese open models catch the US frontier? ▪️How much does distillation explain China's progress ▪️Reactions to Anthropic Backlash ▪️What a Stuxnet style backdoor in model weights looks like ▪️The OpenAI and Hugging Face breach ▪️xAI and Cursor ▪️Prediction check-ins 0:00 Intro 2:41 China's Open Source Models Catch Up 7:42 Does Distillation Explain China's Rise? 13:54 The Geopolitical Risk of Chinese AI Models 20:28 Should the Government Restrict Open Models? 22:06 What Are the Labs Really Learning From You? 29:22 Future of Government Regulation 40:50 The OpenAI-Hugging Face Hack 47:59 Grok, Cursor, and the Value of Real Data 56:52 SSI and OpenRouter 1:02:52 Venture Capital's Return to Deep Tech 1:05:16 Quickfire YouTube: Spotify: Apple:

Jacob Effron

14,518 Aufrufe • vor 1 Monat

I had a fantastic time discussing with the learning legend Justin Skycak from Math Academy about learning math in the modern age. we've talked about his quite impressive self-learning journey (3000h of math in high school) all the way to how he hand curated the initial knowledge graph for math academy to make that process more efficient. great lively 3h discussion here are the chapters: 0:00:00 - intro: 0:02:10 - justin background 0:05:45 - 3000h math self study in high school 0:11:45 - what a day looked like for that 3000h stretch 0:16:10 - meta-learning vs pure math learning 0:21:50 - when did you get into cognitive neuro? 0:29:55 - how did the fundamental math helped in your research projects 0:43:10 - what does the math academy learning system looks like 0:47:34 - how did you guys build the 2000 topic knowledge graph 1:01:15 - would LLM be useful as an interface to that knowledge graph for the students? 1:10:46 - how does the FIRe spaced repetition algorithm works? 1:17:34 - does the same knowledge graph structure would work for physics? or other topic?: 1:34:05 - how do you understand the subject vs the curiculum 1:35:50 - is there a connection between studying math and learning a sport? 1:42:00 - do you think in math doing and teaching requires different skills? 1:56:25 - could you get understanding without automaticy? 2:05:35 - do you see any upside of confusion in learning? 2:14:11 - learning math as an adult? 2:19:20 - how to fill the motivation gap after learning the fundamental? 2:24:10 - how should teaching math for kids and adults balance fundamentals and creativity? 2:33:55 - is it ever too late to learn math seriously? 2:46:00 - mastery learning vs ultra learning 2:51:30 - top-down vs bottom-up 2:53:40 - mastery learning for domain without a structured hierarchical structure? 2:56:30 - neurodivergence / adhd for structured math learning? 3:06:20 - amateur mathematician augmented with technology will be able to contribute to research? 3:14:37 - what are you most excited about right now in term of learning enjoy!

I had a fantastic time discussing with the learning legend Justin Skycak from Math Academy about learning math in the modern age. we've talked about his quite impressive self-learning journey (3000h of math in high school) all the way to how he hand curated the initial knowledge graph for math academy to make that process more efficient. great lively 3h discussion here are the chapters: 0:00:00 - intro: 0:02:10 - justin background 0:05:45 - 3000h math self study in high school 0:11:45 - what a day looked like for that 3000h stretch 0:16:10 - meta-learning vs pure math learning 0:21:50 - when did you get into cognitive neuro? 0:29:55 - how did the fundamental math helped in your research projects 0:43:10 - what does the math academy learning system looks like 0:47:34 - how did you guys build the 2000 topic knowledge graph 1:01:15 - would LLM be useful as an interface to that knowledge graph for the students? 1:10:46 - how does the FIRe spaced repetition algorithm works? 1:17:34 - does the same knowledge graph structure would work for physics? or other topic?: 1:34:05 - how do you understand the subject vs the curiculum 1:35:50 - is there a connection between studying math and learning a sport? 1:42:00 - do you think in math doing and teaching requires different skills? 1:56:25 - could you get understanding without automaticy? 2:05:35 - do you see any upside of confusion in learning? 2:14:11 - learning math as an adult? 2:19:20 - how to fill the motivation gap after learning the fundamental? 2:24:10 - how should teaching math for kids and adults balance fundamentals and creativity? 2:33:55 - is it ever too late to learn math seriously? 2:46:00 - mastery learning vs ultra learning 2:51:30 - top-down vs bottom-up 2:53:40 - mastery learning for domain without a structured hierarchical structure? 2:56:30 - neurodivergence / adhd for structured math learning? 3:06:20 - amateur mathematician augmented with technology will be able to contribute to research? 3:14:37 - what are you most excited about right now in term of learning enjoy!

Yacine Mahdid

57,320 Aufrufe • vor 5 Monaten

Emil Michael: America’s Regulation Problems and Opportunities The great innovations of the past 20 years — Uber, Airbnb, Bitcoin — all involved companies dealing with real-world regulation. Emil Michael led this charge at Uber. As AI potentially impacts all industries, self driving cars finally start operating, and the Supreme Court’s reversal of Chevron suggests regulatory reform, what is in store for this next era of innovation in America? Emil and I spent 90 minutes talking about: - Why California’s rail track will cost $128 billion - What self-driving means for rideshare in 2024 - How to fix our regulatory landscape Watch the video here: 00:00 intro 00:42 regulatory capture 03:23 why we can’t build anything 09:25 software culture in America 11:20 “Uber for X” opportunities 17:07 winner-take-most markets 30:45 self driving cars and Uber 43:00 becoming a “talent vortex” 49:20 the New World of PR 58:28 what Emil is doing now 1:03:00 the landscape for founders 1:11:20 our antitrust regime 1:24:30 writing better policy For the full video in beautiful 4K quality, visit YouTube!

Emil Michael: America’s Regulation Problems and Opportunities The great innovations of the past 20 years — Uber, Airbnb, Bitcoin — all involved companies dealing with real-world regulation. Emil Michael led this charge at Uber. As AI potentially impacts all industries, self driving cars finally start operating, and the Supreme Court’s reversal of Chevron suggests regulatory reform, what is in store for this next era of innovation in America? Emil and I spent 90 minutes talking about: - Why California’s rail track will cost $128 billion - What self-driving means for rideshare in 2024 - How to fix our regulatory landscape Watch the video here: 00:00 intro 00:42 regulatory capture 03:23 why we can’t build anything 09:25 software culture in America 11:20 “Uber for X” opportunities 17:07 winner-take-most markets 30:45 self driving cars and Uber 43:00 becoming a “talent vortex” 49:20 the New World of PR 58:28 what Emil is doing now 1:03:00 the landscape for founders 1:11:20 our antitrust regime 1:24:30 writing better policy For the full video in beautiful 4K quality, visit YouTube!

Will Robbins

23,751 Aufrufe • vor 2 Jahren

auto-research is starting to gain traction as a very viable paradigm for creating useful research discovery. now, that paradigm is still in its infancy and the infrastructure to hold all that trail of context as the agents blaze through experiments isn't well defined (to say the least). on that topic, I had the chance to chat with my boys francesco and giulio from paradigma about what underlying infra is needed to make this paradigm work. the paradigma's paradigm, which involves copious amount of DAGs, make this auto-research paradigm a paradigmatic case of essential infrastructure. here's the full video in full: - 0:00 - what is missing from auto-research? - 2:02 - giulio and francesco ai journey - 8:10 - research infra is the bottleneck? - 10:18 - paradigma vision of autonomous research - 13:17 - “important discovery per joules” - 17:15 - why is DAG the unit of research for auto-research? - 20:40 - is paradigma trying to replace the research publication? - 24:50 - how does knowledge is shared between experiments in the DAG? - 27:34 - what is even auto-research lol? - 33:53 - the value of the human mind in this auto-research future. - 37:00 - how do you reconcile hallucination in this auto-research paradigm? - 41:33 - the adoption of auto-research across varied fields? - 47:30 - ✨ introduction to the auto-research infrastructure. ✨ - 56:55 - where is the code? - 59:10 - full IDE next? - 1:03:20 - the place of the human in this DAG / code quality? manual node? token spent? - 1:16:02 - who’s the user for auto-research? - 1:18:13 - how to validate bad DAG? - 1:20:18 - ✨ auto-research agent results ✨ - 1:22:53 - ✨ how a big research DAG looks like? ✨ - 1:25:10 - how to get the canonical DAG for the final result? - 1:27:50 - the auto-research DAG being the new pre-print? - 1:30:05 - what’s next for paradigma and the auto-research infra? - 1:35:00 - what are they excited about research wise? enjoyyyyy my guys 🌹

auto-research is starting to gain traction as a very viable paradigm for creating useful research discovery. now, that paradigm is still in its infancy and the infrastructure to hold all that trail of context as the agents blaze through experiments isn't well defined (to say the least). on that topic, I had the chance to chat with my boys francesco and giulio from paradigma about what underlying infra is needed to make this paradigm work. the paradigma's paradigm, which involves copious amount of DAGs, make this auto-research paradigm a paradigmatic case of essential infrastructure. here's the full video in full: - 0:00 - what is missing from auto-research? - 2:02 - giulio and francesco ai journey - 8:10 - research infra is the bottleneck? - 10:18 - paradigma vision of autonomous research - 13:17 - “important discovery per joules” - 17:15 - why is DAG the unit of research for auto-research? - 20:40 - is paradigma trying to replace the research publication? - 24:50 - how does knowledge is shared between experiments in the DAG? - 27:34 - what is even auto-research lol? - 33:53 - the value of the human mind in this auto-research future. - 37:00 - how do you reconcile hallucination in this auto-research paradigm? - 41:33 - the adoption of auto-research across varied fields? - 47:30 - ✨ introduction to the auto-research infrastructure. ✨ - 56:55 - where is the code? - 59:10 - full IDE next? - 1:03:20 - the place of the human in this DAG / code quality? manual node? token spent? - 1:16:02 - who’s the user for auto-research? - 1:18:13 - how to validate bad DAG? - 1:20:18 - ✨ auto-research agent results ✨ - 1:22:53 - ✨ how a big research DAG looks like? ✨ - 1:25:10 - how to get the canonical DAG for the final result? - 1:27:50 - the auto-research DAG being the new pre-print? - 1:30:05 - what’s next for paradigma and the auto-research infra? - 1:35:00 - what are they excited about research wise? enjoyyyyy my guys 🌹

Yacine Mahdid

12,091 Aufrufe • vor 3 Monaten

In 2016, at an AI conference in NYC, I explained artificial consciousness, world models, predictive coding, and science as data compression in less than 10 minutes. I happened to be in town, walked in without being announced, and ended up on their panel. It was great fun. Organizer: David Chalmers Distinguished panel members: DanielKahneman Susan Schneider Gary Marcus Jaan Tallin Original YouTube video: 29:40-39:15 Two relevant pages from the AI Blog: 1990: planning & reinforcement learning with recurrent world models and artificial curiosity. This yields a simple explanation of consciousness and self-awareness: 1991: first very deep learning with self-supervised pre-training. A conscious chunker recurrent neural net (RNN) attends to unexpected events that surprise a lower-level subconscious automatiser RNN. The automatiser uses neural network distillation to compress and absorb the formerly conscious insights and behaviours of the chunker, thus making them subconscious: Interview: 2016: J. Carmichael. Artificial Intelligence Gained Consciousness in 1991. Why A.I. pioneer Jürgen Schmidhuber is convinced the ultimate breakthrough already happened. Inverse, 2016.

In 2016, at an AI conference in NYC, I explained artificial consciousness, world models, predictive coding, and science as data compression in less than 10 minutes. I happened to be in town, walked in without being announced, and ended up on their panel. It was great fun. Organizer: David Chalmers Distinguished panel members: DanielKahneman Susan Schneider Gary Marcus Jaan Tallin Original YouTube video: 29:40-39:15 Two relevant pages from the AI Blog: 1990: planning & reinforcement learning with recurrent world models and artificial curiosity. This yields a simple explanation of consciousness and self-awareness: 1991: first very deep learning with self-supervised pre-training. A conscious chunker recurrent neural net (RNN) attends to unexpected events that surprise a lower-level subconscious automatiser RNN. The automatiser uses neural network distillation to compress and absorb the formerly conscious insights and behaviours of the chunker, thus making them subconscious: Interview: 2016: J. Carmichael. Artificial Intelligence Gained Consciousness in 1991. Why A.I. pioneer Jürgen Schmidhuber is convinced the ultimate breakthrough already happened. Inverse, 2016.

Jürgen Schmidhuber

327,681 Aufrufe • vor 2 Jahren

An ex-Google engineer just dropped a free 2-hour course on graph engineering and AI agents From prompts to self-improving agents 👇 0% → 00:00 - know when to use loops vs graphs 20% → 20:29 - benchmark 5 memory layers head to head 40% → 49:29 - 8 lessons from AI founders and investors 60% → 1:08:17 - see how a self-improving agent learns 80% → 1:28:12 - wire up tracing and evals 100% → 1:47:15 - understand semantic vs episodic memory Most people still think an agent is just a prompt with tools This shows what actually runs around the model, with open-source code and live demos Worth more than most paid agent courses Bookmark it and watch this weekend 👇

An ex-Google engineer just dropped a free 2-hour course on graph engineering and AI agents From prompts to self-improving agents 👇 0% → 00:00 - know when to use loops vs graphs 20% → 20:29 - benchmark 5 memory layers head to head 40% → 49:29 - 8 lessons from AI founders and investors 60% → 1:08:17 - see how a self-improving agent learns 80% → 1:28:12 - wire up tracing and evals 100% → 1:47:15 - understand semantic vs episodic memory Most people still think an agent is just a prompt with tools This shows what actually runs around the model, with open-source code and live demos Worth more than most paid agent courses Bookmark it and watch this weekend 👇

spect

24,035 Aufrufe • vor 7 Tagen

Charles Broskoski believes everything should be personal: your work, your tools, your taste, and perhaps most provocatively, the business you build. I talked to Charles Broskoski about knowing yourself, noticing patterns, casual research, and building a personal business in Are.na. Charles is an artist turned software engineer and co-founder/CEO of Arena: a platform for collecting, connecting, and self-directed learning. Arena is coming up on its 15th anniversary, and Charles hopes it can last another 15 and far beyond. I discovered him through his essay and talk, "Here for the Wrong Reasons", and was struck by his philosophy that who we are is made up of the things we pay attention to, and specifically what we see in them. We discuss that and more: - Why creativity is decision-making and problem-solving - Charles's case for "casual" research - Why skateboarding is a model for being a beginner - How information that inflects you becomes part of who you are - Personal versus performative taste (and why it shouldn't be thought of as a competitive skill) - Creating a typeface to replace Arial that looks (almost) exactly like Arial - Why more creative people should start businesses - Drawing inspiration from a 1,300 year old Japanese hot spring hotel - "The reward is the work. That you get to hang out with your friends and make something cool." - Why Arena wants to fade into the background Timestamps: 0:00 - Opening Highlights 1:21 - Intro: Charles Broskoski 4:00 - Thanks to Notion 5:26 - Start: Creativity as Self-Knowledge and Problem-Solving 13:37 - Self-directed Learning and Casual Research 21:33 - Skateboarding, Being a Beginner, In Defense of Posers 33:26 - Contextual Patterns and Channels 45:54 - Nodal Points, Your Radar, and Careful Attention 1:04:57 - Subjectivity, Self-Knowledge, and Taste 1:15:09 - Performance: Here for Fame and Not Love 1:22:53 - Aspirational Attention 1:29:02 - Designing Generous Tools 1:42:44 - Space in a Product and Fading into the Background 1:50:01 - Why Creatives Should Be Entrepreneurial & Building an Independent Business Online 1:54:11 - Patience, Durability, and Antifragility 1:59:48 - Personal Businesses 2:10:27 - Grab Bag: Authenticity, Bohm Dialogue, Skateboarding, and Keeping Things Personal 2:28:28 - Thanks Again to Notion Dialectic with Jackson Dahl Ep. 40: Charles Broskoski - Everything is Personal - is out now, below and on all platforms.

Charles Broskoski believes everything should be personal: your work, your tools, your taste, and perhaps most provocatively, the business you build. I talked to Charles Broskoski about knowing yourself, noticing patterns, casual research, and building a personal business in Are.na. Charles is an artist turned software engineer and co-founder/CEO of Arena: a platform for collecting, connecting, and self-directed learning. Arena is coming up on its 15th anniversary, and Charles hopes it can last another 15 and far beyond. I discovered him through his essay and talk, "Here for the Wrong Reasons", and was struck by his philosophy that who we are is made up of the things we pay attention to, and specifically what we see in them. We discuss that and more: - Why creativity is decision-making and problem-solving - Charles's case for "casual" research - Why skateboarding is a model for being a beginner - How information that inflects you becomes part of who you are - Personal versus performative taste (and why it shouldn't be thought of as a competitive skill) - Creating a typeface to replace Arial that looks (almost) exactly like Arial - Why more creative people should start businesses - Drawing inspiration from a 1,300 year old Japanese hot spring hotel - "The reward is the work. That you get to hang out with your friends and make something cool." - Why Arena wants to fade into the background Timestamps: 0:00 - Opening Highlights 1:21 - Intro: Charles Broskoski 4:00 - Thanks to Notion 5:26 - Start: Creativity as Self-Knowledge and Problem-Solving 13:37 - Self-directed Learning and Casual Research 21:33 - Skateboarding, Being a Beginner, In Defense of Posers 33:26 - Contextual Patterns and Channels 45:54 - Nodal Points, Your Radar, and Careful Attention 1:04:57 - Subjectivity, Self-Knowledge, and Taste 1:15:09 - Performance: Here for Fame and Not Love 1:22:53 - Aspirational Attention 1:29:02 - Designing Generous Tools 1:42:44 - Space in a Product and Fading into the Background 1:50:01 - Why Creatives Should Be Entrepreneurial & Building an Independent Business Online 1:54:11 - Patience, Durability, and Antifragility 1:59:48 - Personal Businesses 2:10:27 - Grab Bag: Authenticity, Bohm Dialogue, Skateboarding, and Keeping Things Personal 2:28:28 - Thanks Again to Notion Dialectic with Jackson Dahl Ep. 40: Charles Broskoski - Everything is Personal - is out now, below and on all platforms.

Jackson Dahl

27,561 Aufrufe • vor 6 Monaten

Inside Nemotron and NVIDIA's AI lab: my conversation with Bryan Catanzaro (Bryan Catanzaro). NVIDIA is a chip company. So why does it put hundreds of researchers on building AI models - and then give them away for free? We go deep into the Nemotron models, what it takes to build a top AI lab, and the future of frontier AI. 01:33 - Is open source AI catching the frontier? 05:29 - Do closed labs blocking distillation slow open source down? 07:42 - Is the US falling behind China? 10:30 - Why companies actually choose open models 12:39 - A "crazy" 2008 bet: machine learning on GPUs 15:33 - Working with Andrew Ng and Dario Amodei at Baidu 17:41 - Coming back to NVIDIA: DLSS and the birth of Megatron 21:55 - The real reason NVIDIA builds its own models 24:28 - Is Moore's Law really dead? 33:37 - The Nemotron family: Nano, Super, Ultra 35:09 - Built for agents: why NVIDIA bets on speed 36:02 - How you train a 550B model in 4 bits 39:25 - Hybrid Mamba-Transformer, explained simply 42:31 - Mixture of experts, and why NVIDIA built NVL72 around it 47:26 - Why a 1-million-token context window matters 49:26 - Multi-token prediction: how the model predicts 5 tokens at once 52:47 - Multi-teacher distillation: teaching one model from many 58:01 - Where reinforcement learning goes next 01:00:16 - Inside NVIDIA's research org: "the mission is the boss" 01:04:03 - How NVIDIA decides who gets the GPUs 01:10:53 - Why NVIDIA still feels entrepreneurial after 33 years 01:12:58 - Why Bryan doesn't believe in the singularity 01:17:50 - The AI backlash 01:19:18 - The controversial case: open AI is safer than closed

Inside Nemotron and NVIDIA's AI lab: my conversation with Bryan Catanzaro (Bryan Catanzaro). NVIDIA is a chip company. So why does it put hundreds of researchers on building AI models - and then give them away for free? We go deep into the Nemotron models, what it takes to build a top AI lab, and the future of frontier AI. 01:33 - Is open source AI catching the frontier? 05:29 - Do closed labs blocking distillation slow open source down? 07:42 - Is the US falling behind China? 10:30 - Why companies actually choose open models 12:39 - A "crazy" 2008 bet: machine learning on GPUs 15:33 - Working with Andrew Ng and Dario Amodei at Baidu 17:41 - Coming back to NVIDIA: DLSS and the birth of Megatron 21:55 - The real reason NVIDIA builds its own models 24:28 - Is Moore's Law really dead? 33:37 - The Nemotron family: Nano, Super, Ultra 35:09 - Built for agents: why NVIDIA bets on speed 36:02 - How you train a 550B model in 4 bits 39:25 - Hybrid Mamba-Transformer, explained simply 42:31 - Mixture of experts, and why NVIDIA built NVL72 around it 47:26 - Why a 1-million-token context window matters 49:26 - Multi-token prediction: how the model predicts 5 tokens at once 52:47 - Multi-teacher distillation: teaching one model from many 58:01 - Where reinforcement learning goes next 01:00:16 - Inside NVIDIA's research org: "the mission is the boss" 01:04:03 - How NVIDIA decides who gets the GPUs 01:10:53 - Why NVIDIA still feels entrepreneurial after 33 years 01:12:58 - Why Bryan doesn't believe in the singularity 01:17:50 - The AI backlash 01:19:18 - The controversial case: open AI is safer than closed

Matt Turck

56,954 Aufrufe • vor 2 Monaten

New podcast with finbarr! We survey the latest post-training recipes, from GLM 5.1, Kimi K2.6, DeepSeek V4, Xiaomi MiMo V2.5, Nemotron Ultra, etc. and discuss: - Why the industry slowly shifted to multi-teacher on-policy distillation (MOPD). - What an Olmo-style recipe would need improvements in - How post-training works / suits larger organizational efforts - Career advice in the foothills of the singularity - and other topics I heard y'all wanted me to start doing this, so making some time when I'm in funemployment! Chapters: 00:00 Introduction & Olmo reflections 06:28 Post-train recipes review (history) 23:00 2026’s model recipes (MiMo Flash, DeepSeek V4, GLM 5, Kimi K2.6, etc.) 39:05 Open-ended post-training discussions 48:22 Career advice in the LLM race Links below, please follow Interconnects AI and like and subscribe and buy my book?

New podcast with finbarr! We survey the latest post-training recipes, from GLM 5.1, Kimi K2.6, DeepSeek V4, Xiaomi MiMo V2.5, Nemotron Ultra, etc. and discuss: - Why the industry slowly shifted to multi-teacher on-policy distillation (MOPD). - What an Olmo-style recipe would need improvements in - How post-training works / suits larger organizational efforts - Career advice in the foothills of the singularity - and other topics I heard y'all wanted me to start doing this, so making some time when I'm in funemployment! Chapters: 00:00 Introduction & Olmo reflections 06:28 Post-train recipes review (history) 23:00 2026’s model recipes (MiMo Flash, DeepSeek V4, GLM 5, Kimi K2.6, etc.) 39:05 Open-ended post-training discussions 48:22 Career advice in the LLM race Links below, please follow Interconnects AI and like and subscribe and buy my book?

Nathan Lambert

43,256 Aufrufe • vor 3 Monaten

How GPT-5 thinks, with OpenAI VP of Research Jerry Tworek 00:00 - Intro 01:01 - What Reasoning Actually Means in AI 02:32 - Chain of Thought: Models Thinking in Words 05:25 - How Models Decide How Long to Think 07:24 - Evolution from o1 to o3 to GPT-5 11:00 - The Road to OpenAI: Growing up in Poland, Dropping out of School, Trading 20:32 - Working on Robotics and Rubik's Cube Solving 23:02 - A Day in the Life: Talking to Researchers 24:06 - How Research Priorities Are Determined 26:53 - OpenAI's Culture of Transparency 29:32 - Balancing Research with Shipping Fast 31:52 - Using OpenAI's Own Tools Daily 32:43 - Pre-Training Plus RL: The Modern AI Stack 35:10 - Reinforcement Learning 101: Training Dogs 40:17 - The Evolution of Deep Reinforcement Learning 42:09 - When GPT-4 Seemed Underwhelming at First 45:39 - How RLHF Made GPT-4 Actually Useful 48:02 - Unsupervised vs Supervised Learning 49:59 - GRPO and How DeepSeek Accelerated US Research 53:05 - What It Takes to Scale Reinforcement Learning 55:36 - Agentic AI and Long-Horizon Thinking 59:19 - Alignment as an RL Problem 1:01:11 - Winning ICPC World Finals Without Specific Training 1:05:53 - Applying RL Beyond Math and Coding 1:09:15 - The Path from Here to AGI 1:12:23 - Pure RL vs Language Models

How GPT-5 thinks, with OpenAI VP of Research Jerry Tworek 00:00 - Intro 01:01 - What Reasoning Actually Means in AI 02:32 - Chain of Thought: Models Thinking in Words 05:25 - How Models Decide How Long to Think 07:24 - Evolution from o1 to o3 to GPT-5 11:00 - The Road to OpenAI: Growing up in Poland, Dropping out of School, Trading 20:32 - Working on Robotics and Rubik's Cube Solving 23:02 - A Day in the Life: Talking to Researchers 24:06 - How Research Priorities Are Determined 26:53 - OpenAI's Culture of Transparency 29:32 - Balancing Research with Shipping Fast 31:52 - Using OpenAI's Own Tools Daily 32:43 - Pre-Training Plus RL: The Modern AI Stack 35:10 - Reinforcement Learning 101: Training Dogs 40:17 - The Evolution of Deep Reinforcement Learning 42:09 - When GPT-4 Seemed Underwhelming at First 45:39 - How RLHF Made GPT-4 Actually Useful 48:02 - Unsupervised vs Supervised Learning 49:59 - GRPO and How DeepSeek Accelerated US Research 53:05 - What It Takes to Scale Reinforcement Learning 55:36 - Agentic AI and Long-Horizon Thinking 59:19 - Alignment as an RL Problem 1:01:11 - Winning ICPC World Finals Without Specific Training 1:05:53 - Applying RL Beyond Math and Coding 1:09:15 - The Path from Here to AGI 1:12:23 - Pure RL vs Language Models

Matt Turck

451,229 Aufrufe • vor 11 Monaten

Thanksgiving-week treat: an epic conversation on Frontier AI with Lukasz Kaiser -co-author of “Attention Is All You Need” (Transformers) and leading research scientist at OpenAI working on GPT-5.1-era reasoning models. 00:00 – Cold open and intro 01:29 – “AI slowdown” vs a wild week of new frontier models 08:03 – Low-hanging fruit, infra, RL training and better data 11:39 – What is a reasoning model, in plain language 17:02 – Chain-of-thought and training the thinking process with RL 21:39 – Łukasz’s path: from logic and France to Google and Kurzweil 24:20 – Inside the Transformer story and what “attention” really means 28:42 – From Google Brain to OpenAI: culture, scale and GPUs 32:49 – What’s next for pre-training, GPUs and distillation 37:29 – Can we still understand these models? Circuits, sparsity and black boxes 39:42 – GPT-4 → GPT-5 → GPT-5.1: what actually changed 42:40 – Post-training, safety and teaching GPT-5.1 different tones 46:16 – How long should GPT-5.1 think? Reasoning tokens and jagged abilities 47:43 – The five-year-old’s dot puzzle that still breaks frontier models 52:22 – Generalization, child-like learning and whether reasoning is enough 53:48 – Beyond Transformers: ARC, LeCun’s ideas and multimodal bottlenecks 56:10 – GPT-5.1 Codex Max, long-running agents and compaction 1:00:06 – Will foundation models eat most apps? The translation analogy and trust 1:02:34 – What still needs to be solved, and where AI might go next

Thanksgiving-week treat: an epic conversation on Frontier AI with Lukasz Kaiser -co-author of “Attention Is All You Need” (Transformers) and leading research scientist at OpenAI working on GPT-5.1-era reasoning models. 00:00 – Cold open and intro 01:29 – “AI slowdown” vs a wild week of new frontier models 08:03 – Low-hanging fruit, infra, RL training and better data 11:39 – What is a reasoning model, in plain language 17:02 – Chain-of-thought and training the thinking process with RL 21:39 – Łukasz’s path: from logic and France to Google and Kurzweil 24:20 – Inside the Transformer story and what “attention” really means 28:42 – From Google Brain to OpenAI: culture, scale and GPUs 32:49 – What’s next for pre-training, GPUs and distillation 37:29 – Can we still understand these models? Circuits, sparsity and black boxes 39:42 – GPT-4 → GPT-5 → GPT-5.1: what actually changed 42:40 – Post-training, safety and teaching GPT-5.1 different tones 46:16 – How long should GPT-5.1 think? Reasoning tokens and jagged abilities 47:43 – The five-year-old’s dot puzzle that still breaks frontier models 52:22 – Generalization, child-like learning and whether reasoning is enough 53:48 – Beyond Transformers: ARC, LeCun’s ideas and multimodal bottlenecks 56:10 – GPT-5.1 Codex Max, long-running agents and compaction 1:00:06 – Will foundation models eat most apps? The translation analogy and trust 1:02:34 – What still needs to be solved, and where AI might go next

Matt Turck

168,007 Aufrufe • vor 9 Monaten

Here's my conversation with Lucas Atkins and the team at Arcee.ai on their path to training and releasing Trinity Large today. From going all in on open models built end to end in the US 6 months ago to having the model in hand is no easy feet. I loved this conversation on how to design a startup around open models and take a bold step to scale it up. I'm openly an Arcee fan, watching them take risk and pull it off. We discuss: - The state (and future) of open vs. closed models, - The business of selling open models for on-prem deployments, - The story of Arcee AI & going “all-in” on this training run, - The ATOM project, - Building frontier model training teams in 6 months, - and other great topics. I really loved this one, and think you well too. Chapters: 00:00:00 Intro: Arcee AI, Trinity Models & Trinity Large 00:08:26 Transitioning a Company to Pre-training 00:13:00 Technical Decisions: Muon and MoE 00:18:41 Scaling and MoE Training Pain 00:23:14 Post-training and RL Strategies 00:28:09 Team Structure and Data Scaling 00:31:31 The Trinity Manifesto: US Open Weights 00:42:31 Specialized Models and Distillation 00:47:12 Infrastructure and Hosting 400B 00:50:53 Open Source as a Business Moat 00:56:31 Predictions: Best Model in 2026 01:02:29 Lightning Round & Conclusions More great open model builder podcasts coming soon!

Here's my conversation with Lucas Atkins and the team at Arcee.ai on their path to training and releasing Trinity Large today. From going all in on open models built end to end in the US 6 months ago to having the model in hand is no easy feet. I loved this conversation on how to design a startup around open models and take a bold step to scale it up. I'm openly an Arcee fan, watching them take risk and pull it off. We discuss: - The state (and future) of open vs. closed models, - The business of selling open models for on-prem deployments, - The story of Arcee AI & going “all-in” on this training run, - The ATOM project, - Building frontier model training teams in 6 months, - and other great topics. I really loved this one, and think you well too. Chapters: 00:00:00 Intro: Arcee AI, Trinity Models & Trinity Large 00:08:26 Transitioning a Company to Pre-training 00:13:00 Technical Decisions: Muon and MoE 00:18:41 Scaling and MoE Training Pain 00:23:14 Post-training and RL Strategies 00:28:09 Team Structure and Data Scaling 00:31:31 The Trinity Manifesto: US Open Weights 00:42:31 Specialized Models and Distillation 00:47:12 Infrastructure and Hosting 400B 00:50:53 Open Source as a Business Moat 00:56:31 Predictions: Best Model in 2026 01:02:29 Lightning Round & Conclusions More great open model builder podcasts coming soon!

Nathan Lambert

26,872 Aufrufe • vor 7 Monaten

Loved this 22-minute talk on continual learning for AI agents. Must watch for anyone looking to get agents performant and into production. Credit: Soheil Feizi at AI Engineer • Agent learning can happen at three layers: the model (weights), the harness (prompts, tools, skills, code, workflows), and memory (session or persistent). • Two fundamental challenges: (1) getting feedback, meaning how do we know if the agent did well and what it should have done instead, and (2) acting on that feedback, meaning deciding which layer or component to change and how. • Feedback sources differ by stage: In development you have benchmarks with evaluators that score pass/fail. In production you only have logs, which can be judged either automatically (LLMs or code analyzing the log, which is scalable) or by human experts (low volume but critical domain knowledge). • Logs plus feedback aren't enough because they're not testable: A single log with feedback is one observation of what happened. You need to lift it into a replayable learning environment, a simulation with tools, users, and defined evaluators, so candidate fixes can be run, verified, and compared. • Three ways to optimize the agent, with tradeoffs: Model-layer updates (SFT, RL post-training like DPO/GRPO, LoRA) are expensive and need benchmarks and evaluators. Harness updates (trace-to-harness coding agents, prompt search like GEPA) are flexible but either untestable and "vibe-based" or benchmark-dependent. Memory updates (fact storage like Letta/Mem0, skill distillation) are cheapest and fastest but usually unverified. • A good learning engine makes "the smallest durable change at the right layer" of the agent. • Verifiable continual learning (VCL): Improve an agent from its own experience where every fix is proven to help and proven to break nothing that already worked. It requires an executable test (replayable failure), a measured delta (score before and after), and regression tests (prior tests still pass). • Four principles of practical VCL: Replayability (turn one-off failures into rerunnable tests), holisticness (one failure can have causes in memory, prompts, tools, workflow, or model, so route the fix to the right layer), lifelongness (fix new failures subject to no regression on past environments, with regression handled inside the optimization loop rather than post-hoc), and efficiency (the loop must run frequently and cheaply, without scaling linearly as past environments accumulate). • Three takeaways: (1) Agent continual learning isn't necessarily fine-tuning; many useful updates live in the harness and memory layers. (2) Production logs are not learning environments and must be transformed into replayable ones. (3) The frontier is regression-aware improvement: fixing new failures while verifying you don't break old ones.

Loved this 22-minute talk on continual learning for AI agents. Must watch for anyone looking to get agents performant and into production. Credit: Soheil Feizi at AI Engineer • Agent learning can happen at three layers: the model (weights), the harness (prompts, tools, skills, code, workflows), and memory (session or persistent). • Two fundamental challenges: (1) getting feedback, meaning how do we know if the agent did well and what it should have done instead, and (2) acting on that feedback, meaning deciding which layer or component to change and how. • Feedback sources differ by stage: In development you have benchmarks with evaluators that score pass/fail. In production you only have logs, which can be judged either automatically (LLMs or code analyzing the log, which is scalable) or by human experts (low volume but critical domain knowledge). • Logs plus feedback aren't enough because they're not testable: A single log with feedback is one observation of what happened. You need to lift it into a replayable learning environment, a simulation with tools, users, and defined evaluators, so candidate fixes can be run, verified, and compared. • Three ways to optimize the agent, with tradeoffs: Model-layer updates (SFT, RL post-training like DPO/GRPO, LoRA) are expensive and need benchmarks and evaluators. Harness updates (trace-to-harness coding agents, prompt search like GEPA) are flexible but either untestable and "vibe-based" or benchmark-dependent. Memory updates (fact storage like Letta/Mem0, skill distillation) are cheapest and fastest but usually unverified. • A good learning engine makes "the smallest durable change at the right layer" of the agent. • Verifiable continual learning (VCL): Improve an agent from its own experience where every fix is proven to help and proven to break nothing that already worked. It requires an executable test (replayable failure), a measured delta (score before and after), and regression tests (prior tests still pass). • Four principles of practical VCL: Replayability (turn one-off failures into rerunnable tests), holisticness (one failure can have causes in memory, prompts, tools, workflow, or model, so route the fix to the right layer), lifelongness (fix new failures subject to no regression on past environments, with regression handled inside the optimization loop rather than post-hoc), and efficiency (the loop must run frequently and cheaply, without scaling linearly as past environments accumulate). • Three takeaways: (1) Agent continual learning isn't necessarily fine-tuning; many useful updates live in the harness and memory layers. (2) Production logs are not learning environments and must be transformed into replayable ones. (3) The frontier is regression-aware improvement: fixing new failures while verifying you don't break old ones.

Alex Lieberman

20,417 Aufrufe • vor 2 Monaten