Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Another quick lecture -- I've been asked many times for prereq's to my book and what you should know, so built a little lecture (with GLM 5.2) to cover some more basics. Topics include: 00:00 Introduction & Course Prerequisites 01:37 Language Models Overview 02:47 The LM Head 04:29 Softmax... & Log-Probabilities 06:13 Anatomy of an LM Training Example 06:37 Computing LLM Probabilities (+Phoebe the Dog) 09:52 Three Common Masks in Post-Training 11:03 A Small Decoding Review 12:14 Training an LM: Cross-Entropy 13:23 Optimization & Fine-Tuning 13:55 Pretraining to Midtraining to SFT Pipeline 15:25 Probability Essentials: KL Divergence & Entropy 19:36 Sigmoid & Pairwise Likelihood 20:29 Reinforcement Learning Framing (MDP) 22:28 Transitioning Tools into Post-Training 23:12 Recommended Resources & Wrap-Up  Happy learning and I'm still taking questions from during the course for Q&A videos.show more

Nathan Lambert

102,668 subscribers

52,668 views • 3 months ago •via X (Twitter)

Education Health & Wellness Science & Technology

Anya Rossi• Live Now

Private livecam show

15 Comments

Nathan Lambert3 months ago

Lecture 0:

Nathan Lambert3 months ago

(video is ft. phoebe, second half with her on my lap)

Zach Mueller3 months ago

I’ve really gotta sit down and spend some time with these. Btw is there a space you’ve made yet for students taking the videos to yap? (Discord etc)

Nathan Lambert3 months ago

Yeah there’s a book discord, link on site / description (don’t past discord links on x or get got)

Zach Mueller3 months ago

Bootiful

Juergentron90003 months ago

5.2 has been blowing my socks off. I've never considered selling a kidney for a Mac studio to run it locally. Until now.

kim3 months ago

thank you so much. learning A LOT from everything you’re sharing: from to rlhf book.

Dipanshu Kushwaha3 months ago

That sounds super helpful! It’s great to see you breaking down the essentials for everyone. Can’t wait to dive into it!

Vishal3 months ago

This is amazing , thanks 🤩

veloX3 months ago

KL divergence ve entropy başlığında bir detay: ikisi de bilgi teorisinin temeli ama çoğu kişi entropy'yi "belirsizlik" olarak, KL divergence'ı ise "mesafe" olarak düşünüyor. Aslında ikisi de olasılık dağılımlarının "ekonomisi" hakkında konuşuyor — entropy kaynağın ne kadar "pahalı" olduğunu, KL divergence ise yanlış modelle kodlama yaparken ne kadar "harcadığını" ölçüyor. Post-training'de bu fark kritik: cross-entropy sadece loss, KL divergence ise modelin orijinal bilgisinden ne kadar sapmaya izin verdiğini kontrol ediyor.

Verma3 months ago

Great Thanks for your effort and posting this prerequisite, Nathan!

Eren | AI x Markets3 months ago

If it does one annoying thing for me every day, I am listening.

Dr. Xi Zeng3 months ago

The timestamped prereq lecture is a good filter. If someone cannot point to the LM head screen in the video, I trust their agent product claim a little less.

aitization 𝕏 3 months ago

Interesting

Akshat B3 months ago

Is the LM head really the biggest bottleneck for training efficiency now? I am reading that the softmax bottleneck in that last layer can suppress 95% of the gradient norm during pretraining. It is good you covered this because it makes some patterns almost impossible to learn.

Related Videos

Engineers need to communicate effectively when building AI Systems. These terms will help you use a shared vocabulary. This is useful when discussing concepts, reading papers, or collaborating with teammates. Listed in order. 00:00 Agenda 00:28 1. Large Language Model 01:28 2. Tokenization 02:53 3. Vectorization 04:15 4. Attention 07:22 5. Self-Supervised Learning 12:07 6. Transformer 14:32 7. Fine-tuning 17:05 8. Few-shot Prompting 18:11 9. Retrieval Augmented Generation 20:33 10. Vector Database 23:03 11. Model Context Protocol 25:43 12. Context Engineering 28:17 13. Agents 29:19 14. Reinforcement Learning 34:42 15. Chain of Thought 35:55 16. Reasoning Models 36:36 17. Multi-modal Models 38:21 18. Small Language Models 40:24 19. Distillation 41:47 20. Quantization If you are a software engineer looking to transition to AI, click the link below. AI Engineering Course: #AI #SoftwareEngineering #Agents

Engineers need to communicate effectively when building AI Systems. These terms will help you use a shared vocabulary. This is useful when discussing concepts, reading papers, or collaborating with teammates. Listed in order. 00:00 Agenda 00:28 1. Large Language Model 01:28 2. Tokenization 02:53 3. Vectorization 04:15 4. Attention 07:22 5. Self-Supervised Learning 12:07 6. Transformer 14:32 7. Fine-tuning 17:05 8. Few-shot Prompting 18:11 9. Retrieval Augmented Generation 20:33 10. Vector Database 23:03 11. Model Context Protocol 25:43 12. Context Engineering 28:17 13. Agents 29:19 14. Reinforcement Learning 34:42 15. Chain of Thought 35:55 16. Reasoning Models 36:36 17. Multi-modal Models 38:21 18. Small Language Models 40:24 19. Distillation 41:47 20. Quantization If you are a software engineer looking to transition to AI, click the link below. AI Engineering Course: #AI #SoftwareEngineering #Agents

Gaurav Sen

80,434 views • 1 year ago

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 views • 11 months ago

New Course: Post-training of LLMs Learn to post-train and customize an LLM in this short course, taught by Banghua Zhu, Assistant Professor at the University of Washington University of Washington, and co-founder of @NexusflowX. Training an LLM to follow instructions or answer questions has two key stages: pre-training and post-training. In pre-training, it learns to predict the next word or token from large amounts of unlabeled text. In post-training, it learns useful behaviors such as following instructions, tool use, and reasoning. Post-training transforms a general-purpose token predictor—trained on trillions of unlabeled text tokens—into an assistant that follows instructions and performs specific tasks. Because it is much cheaper than pre-training, it is practical for many more teams to incorporate post-training methods into their workflows than pre-training. In this course, you’ll learn three common post-training methods—Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning (RL)—and how to use each one effectively. With SFT, you train the model on pairs of input and ideal output responses. With DPO, you provide both a preferred (chosen) and a less preferred (rejected) response and train the model to favor the preferred output. With RL, the model generates an output, receives a reward score based on human or automated feedback, and updates the model to improve performance. You’ll learn the basic concepts, common use cases, and principles for curating high-quality data for effective training. Through hands-on labs, you’ll download a pre-trained model from Hugging Face and post-train it using SFT, DPO, and RL to see how each technique shapes model behavior. In detail, you’ll: - Understand what post-training is, when to use it, and how it differs from pre-training. - Build an SFT pipeline to turn a base model into an instruct model. - Explore how DPO reshapes behavior by minimizing contrastive loss—penalizing poor responses and reinforcing preferred ones. - Implement a DPO pipeline to change the identity of a chat assistant. - Learn online RL methods such as Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO), and how to design reward functions. - Train a model with GRPO to improve its math capabilities using a verifiable reward. Post-training is one of the most rapidly developing areas of LLM training. Whether you’re building a high-accuracy context-specific assistant, fine-tuning a model's tone, or improving task-specific accuracy, this course will give you experience with the most important techniques shaping how LLMs are post-trained today. Please sign up here:

New Course: Post-training of LLMs Learn to post-train and customize an LLM in this short course, taught by Banghua Zhu, Assistant Professor at the University of Washington University of Washington, and co-founder of @NexusflowX. Training an LLM to follow instructions or answer questions has two key stages: pre-training and post-training. In pre-training, it learns to predict the next word or token from large amounts of unlabeled text. In post-training, it learns useful behaviors such as following instructions, tool use, and reasoning. Post-training transforms a general-purpose token predictor—trained on trillions of unlabeled text tokens—into an assistant that follows instructions and performs specific tasks. Because it is much cheaper than pre-training, it is practical for many more teams to incorporate post-training methods into their workflows than pre-training. In this course, you’ll learn three common post-training methods—Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Online Reinforcement Learning (RL)—and how to use each one effectively. With SFT, you train the model on pairs of input and ideal output responses. With DPO, you provide both a preferred (chosen) and a less preferred (rejected) response and train the model to favor the preferred output. With RL, the model generates an output, receives a reward score based on human or automated feedback, and updates the model to improve performance. You’ll learn the basic concepts, common use cases, and principles for curating high-quality data for effective training. Through hands-on labs, you’ll download a pre-trained model from Hugging Face and post-train it using SFT, DPO, and RL to see how each technique shapes model behavior. In detail, you’ll: - Understand what post-training is, when to use it, and how it differs from pre-training. - Build an SFT pipeline to turn a base model into an instruct model. - Explore how DPO reshapes behavior by minimizing contrastive loss—penalizing poor responses and reinforcing preferred ones. - Implement a DPO pipeline to change the identity of a chat assistant. - Learn online RL methods such as Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO), and how to design reward functions. - Train a model with GRPO to improve its math capabilities using a verifiable reward. Post-training is one of the most rapidly developing areas of LLM training. Whether you’re building a high-accuracy context-specific assistant, fine-tuning a model's tone, or improving task-specific accuracy, this course will give you experience with the most important techniques shaping how LLMs are post-trained today. Please sign up here:

Andrew Ng

125,146 views • 1 year ago

When should you start post-training your own models? Fireworks CEO Lin Qiao’s answer: after product-market fit. Not because it's hard... but because only after PMF is the data coming off your product surface worth training on. Lin joined us for our Sequoia Capital "Own Your Intelligence" event to host a workshop on all things post-training; what works, what breaks, and how not to let the model outsmart you. Must listen!! 00:00 Introduction 00:37 What Fireworks sees across thousands of AI applications 02:47 Off-the-shelf APIs and the problem of keeping your taste 03:58 What "owning your intelligence" actually means 05:43 The progression: prompting → RAG → SFT → preferences → RL 07:20 Why this mirrors how humans learn 09:03 Matching the technique to the problem you actually have 10:46 Where teams get stuck: data quality and vibe evals 12:28 Reward hacking: the model that wrote zero lines of code 13:59 Training-to-serving alignment (and why quality drops) 15:55 Post-training in healthcare and security 17:31 From coding to every co-work domain 19:35 Incumbents, cost burden, and not scaling into bankruptcy 21:26 How much control do you want? 23:24 Q&A: What makes a good reward signal 25:00 Q&A: When to start thinking about post-training

When should you start post-training your own models? Fireworks CEO Lin Qiao’s answer: after product-market fit. Not because it's hard... but because only after PMF is the data coming off your product surface worth training on. Lin joined us for our Sequoia Capital "Own Your Intelligence" event to host a workshop on all things post-training; what works, what breaks, and how not to let the model outsmart you. Must listen!! 00:00 Introduction 00:37 What Fireworks sees across thousands of AI applications 02:47 Off-the-shelf APIs and the problem of keeping your taste 03:58 What "owning your intelligence" actually means 05:43 The progression: prompting → RAG → SFT → preferences → RL 07:20 Why this mirrors how humans learn 09:03 Matching the technique to the problem you actually have 10:46 Where teams get stuck: data quality and vibe evals 12:28 Reward hacking: the model that wrote zero lines of code 13:59 Training-to-serving alignment (and why quality drops) 15:55 Post-training in healthcare and security 17:31 From coding to every co-work domain 19:35 Incumbents, cost burden, and not scaling into bankruptcy 21:26 How much control do you want? 23:24 Q&A: What makes a good reward signal 25:00 Q&A: When to start thinking about post-training

Sonya Huang 🐥

28,212 views • 1 month ago

Want world class research capabilities, but don’t have the resources of a big lab? At our recent Sovereign AI event, Gabe Pereyra shared Harvey ’s “moneyball” approach. Here’s the playbook: 00:00 Introduction 00:37 Building a research lab on a budget 02:28 Legal Agent Bench, contracting, and the diligence dataset 03:57 Domain experts guiding synthetic data generation 05:23 Why Harvey open sourced its datasets 06:55 Working with the neo labs – and why more than one 08:20 Post-training in-house: building "Associate 1" 09:44 The model serving matrix: 60 countries, fallbacks, SLAs 11:05 Deciding what stays in production 12:29 Simple open source switches and model routing 13:55 Moneyball: "If we win on this budget, we change the game" 14:53 Q&A: Training with sensitive data 17:16 Q&A: Competing for research talent 18:46 Q&A: Designing rubrics that actually challenge frontier models 20:19 Q&A: Where the pipeline breaks — data, research, or infra 22:59 Q&A: The tension in open sourcing a benchmark 25:02 Q&A: Biggest remaining open problems 27:10 Q&A: Competing with horizontal products

Want world class research capabilities, but don’t have the resources of a big lab? At our recent Sovereign AI event, Gabe Pereyra shared Harvey ’s “moneyball” approach. Here’s the playbook: 00:00 Introduction 00:37 Building a research lab on a budget 02:28 Legal Agent Bench, contracting, and the diligence dataset 03:57 Domain experts guiding synthetic data generation 05:23 Why Harvey open sourced its datasets 06:55 Working with the neo labs – and why more than one 08:20 Post-training in-house: building "Associate 1" 09:44 The model serving matrix: 60 countries, fallbacks, SLAs 11:05 Deciding what stays in production 12:29 Simple open source switches and model routing 13:55 Moneyball: "If we win on this budget, we change the game" 14:53 Q&A: Training with sensitive data 17:16 Q&A: Competing for research talent 18:46 Q&A: Designing rubrics that actually challenge frontier models 20:19 Q&A: Where the pipeline breaks — data, research, or infra 22:59 Q&A: The tension in open sourcing a benchmark 25:02 Q&A: Biggest remaining open problems 27:10 Q&A: Competing with horizontal products

Pat Grady

179,308 views • 1 month ago

#명재현 댄스 챌린지 모음 zip 00:00 Candy 00:26 특 00:51 # menow 01:20 최애의 아이 01:36 NIGHT DANCER 01:57 Seven 02:21 Cream Soda 02:47 RINDAMAN 03:23 In Bloom 03:41 ALOHA 04:12 해금 04:29 Killin' Me Good 04:56 Bubble 05:19 블루체크 05:36 Ready or Not 05:49 Perfume 06:04 겐토 06:29 ビートDEトーヒ 07:00 3D 07:41 Worth It 07:52 B.O.M.B 08:28 첫 눈 08:43 War Cry 08:58 Perfect Night 09:37 Like I Do 09:48 Honey 10:05 DASH 10:27 Standing Next To You 10:46 Hype Boy 11:13 첫 만남은 계획대로 되지 않아 11:41 사랑스러워 12:07 때깔 12:22 Legit 12:51 Water 13:04 Two of Hearts 13:12 EASY 13:32 Smoke Remix 14:09 TAP 14:46 Martini Blue 15:03 BODY 15:26 PICK IT UP 15:41 LIGHTHOUSE 16:00 EENIE MEENIE 16:22 Nectar 16:45 C'est La Vie 17:07 쩔어 17:26 꿈빛 파티시엘 17:56 Impossible 18:14 포켓몬 18:29 SWEAT 18:51 빠나나날라 19:41 SPOT! 20:05 포철고 20:28 빛 20:56 MAESTRO 21:21 Girls Never Die 21:45 KICK IT 22:12 Armageddon 22:39 Taxi Blurr 23:00 Nothin' on You 23:20 Feel the POP 23:39 Shooting Star 24:00 진격의 방탄 24:16 ZOMBIE 24:30 LIKE THAT 24:45 ABCD 25:01 Boom Boom Bass 25:16 Yo Bunny 25:31 BADVILLAIN 25:51 올드스쿨 뉴스쿨 26:08 Supernatural 26:31 How Sweet 27:08 내일에서 기다릴게 27:31 Peaches 27:46 GOOD SO BAD 28:13 별별별 28:41 CRAZY 29:03 슈퍼슈퍼 29:32 껌 29:52 I Need A Girl 30:14 SAD SONG 30:32 네모네모 30:51 옴브리뉴 31:12 반가워, 나의 첫사랑 31:33 혀끝 32:00 Hit the Floor 32:16 기억사탕 32:34 Cherish

#명재현 댄스 챌린지 모음 zip 00:00 Candy 00:26 특 00:51 # menow 01:20 최애의 아이 01:36 NIGHT DANCER 01:57 Seven 02:21 Cream Soda 02:47 RINDAMAN 03:23 In Bloom 03:41 ALOHA 04:12 해금 04:29 Killin' Me Good 04:56 Bubble 05:19 블루체크 05:36 Ready or Not 05:49 Perfume 06:04 겐토 06:29 ビートDEトーヒ 07:00 3D 07:41 Worth It 07:52 B.O.M.B 08:28 첫 눈 08:43 War Cry 08:58 Perfect Night 09:37 Like I Do 09:48 Honey 10:05 DASH 10:27 Standing Next To You 10:46 Hype Boy 11:13 첫 만남은 계획대로 되지 않아 11:41 사랑스러워 12:07 때깔 12:22 Legit 12:51 Water 13:04 Two of Hearts 13:12 EASY 13:32 Smoke Remix 14:09 TAP 14:46 Martini Blue 15:03 BODY 15:26 PICK IT UP 15:41 LIGHTHOUSE 16:00 EENIE MEENIE 16:22 Nectar 16:45 C'est La Vie 17:07 쩔어 17:26 꿈빛 파티시엘 17:56 Impossible 18:14 포켓몬 18:29 SWEAT 18:51 빠나나날라 19:41 SPOT! 20:05 포철고 20:28 빛 20:56 MAESTRO 21:21 Girls Never Die 21:45 KICK IT 22:12 Armageddon 22:39 Taxi Blurr 23:00 Nothin' on You 23:20 Feel the POP 23:39 Shooting Star 24:00 진격의 방탄 24:16 ZOMBIE 24:30 LIKE THAT 24:45 ABCD 25:01 Boom Boom Bass 25:16 Yo Bunny 25:31 BADVILLAIN 25:51 올드스쿨 뉴스쿨 26:08 Supernatural 26:31 How Sweet 27:08 내일에서 기다릴게 27:31 Peaches 27:46 GOOD SO BAD 28:13 별별별 28:41 CRAZY 29:03 슈퍼슈퍼 29:32 껌 29:52 I Need A Girl 30:14 SAD SONG 30:32 네모네모 30:51 옴브리뉴 31:12 반가워, 나의 첫사랑 31:33 혀끝 32:00 Hit the Floor 32:16 기억사탕 32:34 Cherish

먕

91,257 views • 1 year ago

Today, we're joined by Aakanksha Chowdhery, member of technical staff at Reflection, to explore the fundamental shifts required to build true agentic AI. While the industry has largely focused on post-training techniques to improve reasoning, Aakanksha draws on her experience leading pre-training efforts for Google’s PaLM and early Gemini models to argue that pre-training itself must be rethought to move beyond static benchmarks. We explore the limitations of next-token prediction for multi-step workflows and examine how attention mechanisms, loss objectives, and training data must evolve to support long-form reasoning and planning. Aakanksha shares insights on the difference between context retrieval and actual reasoning, the importance of "trajectory" training data, and why scaling remains essential for discovering emergent agentic capabilities like error recovery and dynamic tool learning. 🗒️ For the full list of resources for this episode, visit the show notes page: 📖 CHAPTERS =============================== 00:00 - Introduction 02:26 - Reflection 04:54 - Limitations of post-training for building agents 07:31 - Rethinking pre-training in agents 10:51 - Scaling 11:27 - Evolving attention mechanisms for agentic capabilities 12:39 - Memory as a tool 14:13 - Loss objectives and training data 15:50 - Fine-tuning loss in agent performance 19:37 - Training data 21:29 - Augmenting dominant training data source 24:11 - Overcoming challenges in training on synthetic data 25:47 - Benchmarks 30:44 - Scaling laws in large models versus small models 33:20 - Long-form versus short-form reasoning 37:57 - Agent’s ability to recover from failure 40:15 - Hallucinations and failure recovery 43:53 - Tool use in agents 46:38 - Coding agents 48:37 - How researchers can contribute to agentic AI

Today, we're joined by Aakanksha Chowdhery, member of technical staff at Reflection, to explore the fundamental shifts required to build true agentic AI. While the industry has largely focused on post-training techniques to improve reasoning, Aakanksha draws on her experience leading pre-training efforts for Google’s PaLM and early Gemini models to argue that pre-training itself must be rethought to move beyond static benchmarks. We explore the limitations of next-token prediction for multi-step workflows and examine how attention mechanisms, loss objectives, and training data must evolve to support long-form reasoning and planning. Aakanksha shares insights on the difference between context retrieval and actual reasoning, the importance of "trajectory" training data, and why scaling remains essential for discovering emergent agentic capabilities like error recovery and dynamic tool learning. 🗒️ For the full list of resources for this episode, visit the show notes page: 📖 CHAPTERS =============================== 00:00 - Introduction 02:26 - Reflection 04:54 - Limitations of post-training for building agents 07:31 - Rethinking pre-training in agents 10:51 - Scaling 11:27 - Evolving attention mechanisms for agentic capabilities 12:39 - Memory as a tool 14:13 - Loss objectives and training data 15:50 - Fine-tuning loss in agent performance 19:37 - Training data 21:29 - Augmenting dominant training data source 24:11 - Overcoming challenges in training on synthetic data 25:47 - Benchmarks 30:44 - Scaling laws in large models versus small models 33:20 - Long-form versus short-form reasoning 37:57 - Agent’s ability to recover from failure 40:15 - Hallucinations and failure recovery 43:53 - Tool use in agents 46:38 - Coding agents 48:37 - How researchers can contribute to agentic AI

The TWIML AI Podcast

45,470 views • 9 months ago

📁 2024 #이정현 챌린지 연말결산 00:00 Like I Do (1) 00:19 3D 00:36 흠뻑 00:58 첫 만남은 계획대로 되지 않아 01:22 Grillz 01:41 으르렁 01:55 JACKPOT 02:10 뉴스쿨 올드스쿨 02:37 사랑 없는 노래 02:59 Two of Hearts 03:10 때깔 03:27 전방향 미소년 03:37 전야 03:55 To. X 04:08 my alter ego 04:19 Everytime We Touch 04:34 Donkey Donkey Donk 04:49 What You Won’t Do for Love 04:59 TAP 05:29 Mongkol Challenge 05:51 Boo Thang 06:07 WISH 06:19 올 듯 말 듯 06:45 Smoothie 07:13 Boyfriend 07:29 Swag 07:50 꽁꽁 얼어붙은 한강 위로 08:04 SWEAT 08:25 Earth, Wind & Fire 08:40 포켓몬 댄스 08:53 아낀다 09:06 MILLIONS 09:21 ZOMBIE 09:33 WORK 09:51 Wet & Wild (WATERBOMB) 10:14 LIKE THAT 10:30 내가 S면 넌 나의 N이 되어줘 10:51 Small girl 11:07 STUPIDZ 11:27 Shooting Stars 11:47 Stay With Me 12:06 Treasure 12:20 Open Always Wins 12:45 Underground 13:00 Feels 13:19 Deja vu 13:34 삐그덕 13:48 태양의 후예 14:03 Chk Chk Boom 14:29 다라리 14:35 VOICEMAIL 14:53 Jus Know 15:11 Supernatural 15:36 MCNasty 15:53 Lay It Down 16:09 Loyal 16:33 MILLION DOLLAR BABY 16:49 ZOO 17:17 doodle 17:34 시카노코 (1) 17:45 Spin 17:59 시카노코 (2) 18:10 So Hot 18:30 Big Dawgs 18:44 샹하이 로맨스 19:03 Young, Wild & Free 19:31 Tell Me 19:43 꿈빛 파티시엘 19:59 APT. 20:05 Boyfriend 20:20 Hit the Floor 20:34 혀끝 21:00 LOVE, MONEY, FAME 21:28 I Don’t Care 21:47 TRIGGER 22:04 No Doubt 22:24 머리어깨무릎발 22:34 igloo 22:53 너와의 모든 지금 23:17 Maps 23:34 Knock Knock 23:42 Like I Do (2) 24:04 Santa Tell Me 24:16 Must Have Love #EVNNE #LEEJEONGHYEON #이정현

📁 2024 #이정현 챌린지 연말결산 00:00 Like I Do (1) 00:19 3D 00:36 흠뻑 00:58 첫 만남은 계획대로 되지 않아 01:22 Grillz 01:41 으르렁 01:55 JACKPOT 02:10 뉴스쿨 올드스쿨 02:37 사랑 없는 노래 02:59 Two of Hearts 03:10 때깔 03:27 전방향 미소년 03:37 전야 03:55 To. X 04:08 my alter ego 04:19 Everytime We Touch 04:34 Donkey Donkey Donk 04:49 What You Won’t Do for Love 04:59 TAP 05:29 Mongkol Challenge 05:51 Boo Thang 06:07 WISH 06:19 올 듯 말 듯 06:45 Smoothie 07:13 Boyfriend 07:29 Swag 07:50 꽁꽁 얼어붙은 한강 위로 08:04 SWEAT 08:25 Earth, Wind & Fire 08:40 포켓몬 댄스 08:53 아낀다 09:06 MILLIONS 09:21 ZOMBIE 09:33 WORK 09:51 Wet & Wild (WATERBOMB) 10:14 LIKE THAT 10:30 내가 S면 넌 나의 N이 되어줘 10:51 Small girl 11:07 STUPIDZ 11:27 Shooting Stars 11:47 Stay With Me 12:06 Treasure 12:20 Open Always Wins 12:45 Underground 13:00 Feels 13:19 Deja vu 13:34 삐그덕 13:48 태양의 후예 14:03 Chk Chk Boom 14:29 다라리 14:35 VOICEMAIL 14:53 Jus Know 15:11 Supernatural 15:36 MCNasty 15:53 Lay It Down 16:09 Loyal 16:33 MILLION DOLLAR BABY 16:49 ZOO 17:17 doodle 17:34 시카노코 (1) 17:45 Spin 17:59 시카노코 (2) 18:10 So Hot 18:30 Big Dawgs 18:44 샹하이 로맨스 19:03 Young, Wild & Free 19:31 Tell Me 19:43 꿈빛 파티시엘 19:59 APT. 20:05 Boyfriend 20:20 Hit the Floor 20:34 혀끝 21:00 LOVE, MONEY, FAME 21:28 I Don’t Care 21:47 TRIGGER 22:04 No Doubt 22:24 머리어깨무릎발 22:34 igloo 22:53 너와의 모든 지금 23:17 Maps 23:34 Knock Knock 23:42 Like I Do (2) 24:04 Santa Tell Me 24:16 Must Have Love #EVNNE #LEEJEONGHYEON #이정현

이브..

25,809 views • 1 year ago

The final lecture of my course is an intro to character training! This is a topic that I've been quietly very invested in for ~18 months, as it: * Has potential for high real world impact * Clearly used extensively at frontier labs * Almost no empirical literature exists * More accessible on academic compute This lecture covers what character training is, reviews model specs, constitutions, the differences, the motivations in real world events, some example research papers I like, and open questions in how it relates to post-training/model use generally. Hopefully this brings more people into the field (and reach out if you have questions). It is one of the more research-y chapters in my book, but one that I felt needed the reference. There is still so little, educational content on the topic online. 0:00 Intro 6:22 Part 1: Fundamentals — character, constitutions, and model specs 19:21 Part 2: Character training in practice 23:23 Part 3: Character elicitation without gradient steps 28:03 Part 4: Open questions (and the end of the course) 32:27 The course, complete Thanks for watching. No need to like and subscribe now that the course is done, you definitely wouldn't! h/t to Sharan for leading the technical work I got to do in the space, and Zafir Stojanovski for investing a lot of attention at this book chapter.

The final lecture of my course is an intro to character training! This is a topic that I've been quietly very invested in for ~18 months, as it: * Has potential for high real world impact * Clearly used extensively at frontier labs * Almost no empirical literature exists * More accessible on academic compute This lecture covers what character training is, reviews model specs, constitutions, the differences, the motivations in real world events, some example research papers I like, and open questions in how it relates to post-training/model use generally. Hopefully this brings more people into the field (and reach out if you have questions). It is one of the more research-y chapters in my book, but one that I felt needed the reference. There is still so little, educational content on the topic online. 0:00 Intro 6:22 Part 1: Fundamentals — character, constitutions, and model specs 19:21 Part 2: Character training in practice 23:23 Part 3: Character elicitation without gradient steps 28:03 Part 4: Open questions (and the end of the course) 32:27 The course, complete Thanks for watching. No need to like and subscribe now that the course is done, you definitely wouldn't! h/t to Sharan for leading the technical work I got to do in the space, and Zafir Stojanovski for investing a lot of attention at this book chapter.

Nathan Lambert

57,352 views • 1 month ago

내가 보려고 모은 #조이 역대 브릿지 파트 모음.zip (Ice Cream Cake~Sweet Dreams) 💚 * 레드벨벳 앨범만 포함했습니다 00:00 Somethin Kinda Crazy 00:05 Take It Slow 00:13 사탕 (Candy) 00:20 Dumb Dumb 00:22 Campfire 00:31 Lady’s Room 00:33 Day 1 00:43 Coold World 00:51 My Dear 00:55 Bad Dracula 00:58 Happily Ever After 01:03 빨간 맛 (Red Flavor) 01:09 여름빛 (Mojito) 01:13 I Just 01:17 Kingdom Come 01:23 두 번째 데이트 (My Second Date) 01:27 Attaboy 01:36 Perfect 10 01:42 달빛소리 01:53 Bad Boy 01:59 Time To Love 02:09 #Cookie Jar 02:13 Aitai-tai 02:20 Power Up 02:23 한 여름의 크리스마스 (With You) 02:30 Mr.E 02:35 Mosquito 02:41 RBB (Really Bad Boy) 02:47 So Good 02:51 멋있게 (Sassy Me) 02:54 Swimming Pool 02:57 Sayonara 03:04 짐살라빔 (Zimzalabim) 03:12 Sunny Side Up! 03:16 Milkshake 03:21 친구가 아냐 (Bing Bing) 03:29 안녕, 여름 (Parade) 03:32 LP 03:41 음파음파 (Umpah Umpah) 03:48 카풀 (Carpool) 03:57 Love Is The Way 04:05 Jumpin’ 04:11 눈 맞추고, 손 맞대고 (Eyes Locked, Hand Locked) 04:17 Psycho 04:24 In & Out 04:34 Remember Forever 04:44 Milky Way 04:52 Queendom 05:00 Pose 05:07 Knock On Wood 05:15 Better Be 05:25 Feel My Rhythm 05:29 Rainbow Halo 05:36 Beg For Me 05:43 In My Dreams 05:48 Marionette 05:50 WILDSIDE 05:54 Jackpot 06:00 Snap Snap 06:08 ‘Cause it’s you 06:14 Color of Love 06:16 Birthday 06:19 BYE BYE 06:22 롤러코스터 (On A Ride) 06:29 Zoom 06:40 Celebrate 06:48 Chill Kill 06:56 Knock Knock (Who’s There?) 06:59 Nightmare 07:03 One Kiss 07:10 Wings 07:21 풍경화 (Scenery) 07:32 Sunflower 07:38 Love Arcade 07:44 Bubble 07:51 Night Drive 07:57 Sweet Dreams

내가 보려고 모은 #조이 역대 브릿지 파트 모음.zip (Ice Cream Cake~Sweet Dreams) 💚 * 레드벨벳 앨범만 포함했습니다 00:00 Somethin Kinda Crazy 00:05 Take It Slow 00:13 사탕 (Candy) 00:20 Dumb Dumb 00:22 Campfire 00:31 Lady’s Room 00:33 Day 1 00:43 Coold World 00:51 My Dear 00:55 Bad Dracula 00:58 Happily Ever After 01:03 빨간 맛 (Red Flavor) 01:09 여름빛 (Mojito) 01:13 I Just 01:17 Kingdom Come 01:23 두 번째 데이트 (My Second Date) 01:27 Attaboy 01:36 Perfect 10 01:42 달빛소리 01:53 Bad Boy 01:59 Time To Love 02:09 #Cookie Jar 02:13 Aitai-tai 02:20 Power Up 02:23 한 여름의 크리스마스 (With You) 02:30 Mr.E 02:35 Mosquito 02:41 RBB (Really Bad Boy) 02:47 So Good 02:51 멋있게 (Sassy Me) 02:54 Swimming Pool 02:57 Sayonara 03:04 짐살라빔 (Zimzalabim) 03:12 Sunny Side Up! 03:16 Milkshake 03:21 친구가 아냐 (Bing Bing) 03:29 안녕, 여름 (Parade) 03:32 LP 03:41 음파음파 (Umpah Umpah) 03:48 카풀 (Carpool) 03:57 Love Is The Way 04:05 Jumpin’ 04:11 눈 맞추고, 손 맞대고 (Eyes Locked, Hand Locked) 04:17 Psycho 04:24 In & Out 04:34 Remember Forever 04:44 Milky Way 04:52 Queendom 05:00 Pose 05:07 Knock On Wood 05:15 Better Be 05:25 Feel My Rhythm 05:29 Rainbow Halo 05:36 Beg For Me 05:43 In My Dreams 05:48 Marionette 05:50 WILDSIDE 05:54 Jackpot 06:00 Snap Snap 06:08 ‘Cause it’s you 06:14 Color of Love 06:16 Birthday 06:19 BYE BYE 06:22 롤러코스터 (On A Ride) 06:29 Zoom 06:40 Celebrate 06:48 Chill Kill 06:56 Knock Knock (Who’s There?) 06:59 Nightmare 07:03 One Kiss 07:10 Wings 07:21 풍경화 (Scenery) 07:32 Sunflower 07:38 Love Arcade 07:44 Bubble 07:51 Night Drive 07:57 Sweet Dreams

맑음쨩 🍏

20,678 views • 1 year ago

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 views • 8 months ago

We don't know what most microbial genes do. Can genomic language models help? there's only one way to find out! this is a 1 hour and 42 minute interview with an MIT professor (the famous Yunha Hwang) chatting about these questions, her work in solving them at Tatta Bio, and more. zoomer captions are back too Links in reply! Timestamps: 00:00:00 - Clips + sponsor roll from the wonderful LatchBio 00:02:07 – Introduction 00:02:23 – Why do microbial genomes matter 00:04:07 – Deep learning acceptance in metagenomics 00:05:25 – The case for genomic “context” over sequence matching 00:06:43 – OMG: the only ML-ready metagenomic dataset 00:09:27 – gLM2: A multimodal genomic language model 00:11:06 – What do you do with the output of genomic language models? 00:17:41 – How will OMG evolve? 00:20:26 – Why train on only microbial genomes, as opposed to all genomes? 00:22:58 – Do we need more sequences or more annotations? 00:23:54 – Is there a conserved microbial genome ‘language’? 00:28:11 – What non-obvious things can this genomic language model tell you? 00:33:08 – Semantic deduplication and evaluation 00:37:33 – How does benchmarking work for these types of models? 00:41:31 – Gaia: A genomic search engine 00:44:18 – Even ‘well-studied’ genomes are mostly unannotated 00:50:51 – Using agents on Gaia 00:54:53 – Will genomic language models reshape the tree of life? 00:59:18 – Current limitations of genomic language models 01:08:54 – Directed evolution as training data 01:12:35 – What is Tatta Bio? 01:19:02 – Building Google for genomic sequences (SeqHub) 01:25:46 – How to create communities around scientific OSS 01:29:06 – What’s the purpose in the centralization of the software? 01:35:37 – How will the way science is done change in 10 years?

We don't know what most microbial genes do. Can genomic language models help? there's only one way to find out! this is a 1 hour and 42 minute interview with an MIT professor (the famous Yunha Hwang) chatting about these questions, her work in solving them at Tatta Bio, and more. zoomer captions are back too Links in reply! Timestamps: 00:00:00 - Clips + sponsor roll from the wonderful LatchBio 00:02:07 – Introduction 00:02:23 – Why do microbial genomes matter 00:04:07 – Deep learning acceptance in metagenomics 00:05:25 – The case for genomic “context” over sequence matching 00:06:43 – OMG: the only ML-ready metagenomic dataset 00:09:27 – gLM2: A multimodal genomic language model 00:11:06 – What do you do with the output of genomic language models? 00:17:41 – How will OMG evolve? 00:20:26 – Why train on only microbial genomes, as opposed to all genomes? 00:22:58 – Do we need more sequences or more annotations? 00:23:54 – Is there a conserved microbial genome ‘language’? 00:28:11 – What non-obvious things can this genomic language model tell you? 00:33:08 – Semantic deduplication and evaluation 00:37:33 – How does benchmarking work for these types of models? 00:41:31 – Gaia: A genomic search engine 00:44:18 – Even ‘well-studied’ genomes are mostly unannotated 00:50:51 – Using agents on Gaia 00:54:53 – Will genomic language models reshape the tree of life? 00:59:18 – Current limitations of genomic language models 01:08:54 – Directed evolution as training data 01:12:35 – What is Tatta Bio? 01:19:02 – Building Google for genomic sequences (SeqHub) 01:25:46 – How to create communities around scientific OSS 01:29:06 – What’s the purpose in the centralization of the software? 01:35:37 – How will the way science is done change in 10 years?

owl

44,279 views • 9 months ago

Why AI Progress Suddenly Feels Real - my conversation with Yann Dubois, who co-leads the Post-Training Frontiers team at OpenAI 00:00 - Intro 01:30 - Why recent AI progress feels like a step function 04:13 - Model reliability & the emotional rollercoaster of shipping GPT-5.5 07:33 - How OpenAI structures vertical and horizontal teams 09:49 - Improving model efficiency and test-time compute 12:32 - Yann's journey from Switzerland to OpenAI 15:37 - Reasoning in 2026: Real-world utility vs verifiable rewards 18:34 - GPT-5.5 Thinking vs Pro: Scaling test-time compute 20:09 - How reasoning models become more efficient 23:23 - Pre-training scaling and overcoming the data wall 27:03 - Multimodal data, synthetic data, and embodied AI 31:05 - Demystifying mid-training and post-training 37:21 - Does RL create new capabilities in AI? 38:53 - The challenges and frontier of scaling RL 43:09 - Is building AI models a craft or a strict science 48:21 - How AI models generalize across different domains 54:18 - How reinforcement learning cures AI hallucinations 56:04 - Negative generalization and conflicting instructions 58:05 - Can RL scale to law, medicine, and the broader economy? 1:00:19 - The evaluation bottleneck and Model as a Judge 1:04:21 - Continuous AI progress & continual learning 1:08:49 - Will foundation models eat the agent harness 1:11:23 - Why startups should focus on the last mile of AI

Why AI Progress Suddenly Feels Real - my conversation with Yann Dubois, who co-leads the Post-Training Frontiers team at OpenAI 00:00 - Intro 01:30 - Why recent AI progress feels like a step function 04:13 - Model reliability & the emotional rollercoaster of shipping GPT-5.5 07:33 - How OpenAI structures vertical and horizontal teams 09:49 - Improving model efficiency and test-time compute 12:32 - Yann's journey from Switzerland to OpenAI 15:37 - Reasoning in 2026: Real-world utility vs verifiable rewards 18:34 - GPT-5.5 Thinking vs Pro: Scaling test-time compute 20:09 - How reasoning models become more efficient 23:23 - Pre-training scaling and overcoming the data wall 27:03 - Multimodal data, synthetic data, and embodied AI 31:05 - Demystifying mid-training and post-training 37:21 - Does RL create new capabilities in AI? 38:53 - The challenges and frontier of scaling RL 43:09 - Is building AI models a craft or a strict science 48:21 - How AI models generalize across different domains 54:18 - How reinforcement learning cures AI hallucinations 56:04 - Negative generalization and conflicting instructions 58:05 - Can RL scale to law, medicine, and the broader economy? 1:00:19 - The evaluation bottleneck and Model as a Judge 1:04:21 - Continuous AI progress & continual learning 1:08:49 - Will foundation models eat the agent harness 1:11:23 - Why startups should focus on the last mile of AI

Matt Turck

101,314 views • 4 months ago

New podcast/lecture combo -- a case study in the messy details of Olmo 3 post training & DPO with Scott Geng. It's rare to make time for these discussions, but we cover: What it takes for a research idea to make it into a (near) frontier model. The messy side of DPO (usually data). Organizational challenges in multi-stage post-training recipes. Reflections on where research is heading today, and how people should think of DPO. DPO fundamentals (once more) Other topics. Scott is one of the great student's I've had the pleasure of working with for a few years! Chapters: 00:00 Introduction 04:32 Preference Tuning Basics 08:21 The DPO-Algorithm Zoo 17:06 Scaling Preference Data 21:55 Delta Learning Hypothesis 32:30: Applying it to Olmo 3 36:08 Challenge 1: Data Details 47:17 Challenge 2: Organization Management 51:45 Challenge 3: Scaling Issues 59:15 The Future of Academic Research Excited to keep making this content under the motivation of my book. Really plugging away now :)

New podcast/lecture combo -- a case study in the messy details of Olmo 3 post training & DPO with Scott Geng. It's rare to make time for these discussions, but we cover: What it takes for a research idea to make it into a (near) frontier model. The messy side of DPO (usually data). Organizational challenges in multi-stage post-training recipes. Reflections on where research is heading today, and how people should think of DPO. DPO fundamentals (once more) Other topics. Scott is one of the great student's I've had the pleasure of working with for a few years! Chapters: 00:00 Introduction 04:32 Preference Tuning Basics 08:21 The DPO-Algorithm Zoo 17:06 Scaling Preference Data 21:55 Delta Learning Hypothesis 32:30: Applying it to Olmo 3 36:08 Challenge 1: Data Details 47:17 Challenge 2: Organization Management 51:45 Challenge 3: Scaling Issues 59:15 The Future of Academic Research Excited to keep making this content under the motivation of my book. Really plugging away now :)

Nathan Lambert

24,350 views • 2 months ago

Robots can already fold laundry, make espresso, clean kitchens, and assemble things. The harder problem is getting them to do those tasks reliably, for long periods of time, without a human babysitting them. At Startup School 2026, Physical Intelligence cofounder Chelsea Finn explains what it takes to build general-purpose robots that work in the real world. She shares how reinforcement learning pushed robot throughput up 2x, how their systems can run autonomously for hours, and why she believes robotics is entering its GPT era: moving from specialized models toward general-purpose systems that can work across tasks, robots, and environments. 00:00 — The State of Physical Intelligence 01:23 — What It Takes to Make Robots Useful 05:11 — The Reliability Problem 07:43 — Reinforcement Learning for Robotics 09:35 — Learning From Failures 12:43 — Training Robots to Improve Themselves 14:21 — Can a Robot Work for 13 Hours Straight? 17:36 — Why Robots Need Memory 21:22 — Building a General-Purpose Robot 25:02 — From Fine-Tuning to Out-of-the-Box Models 27:35 — Training on All the Data 30:20 — One Model That Beats the Specialists 31:21 — Compositional Generalization 37:49 — The GPT Era of Robotics 39:49 — Q&A

Robots can already fold laundry, make espresso, clean kitchens, and assemble things. The harder problem is getting them to do those tasks reliably, for long periods of time, without a human babysitting them. At Startup School 2026, Physical Intelligence cofounder Chelsea Finn explains what it takes to build general-purpose robots that work in the real world. She shares how reinforcement learning pushed robot throughput up 2x, how their systems can run autonomously for hours, and why she believes robotics is entering its GPT era: moving from specialized models toward general-purpose systems that can work across tasks, robots, and environments. 00:00 — The State of Physical Intelligence 01:23 — What It Takes to Make Robots Useful 05:11 — The Reliability Problem 07:43 — Reinforcement Learning for Robotics 09:35 — Learning From Failures 12:43 — Training Robots to Improve Themselves 14:21 — Can a Robot Work for 13 Hours Straight? 17:36 — Why Robots Need Memory 21:22 — Building a General-Purpose Robot 25:02 — From Fine-Tuning to Out-of-the-Box Models 27:35 — Training on All the Data 30:20 — One Model That Beats the Specialists 31:21 — Compositional Generalization 37:49 — The GPT Era of Robotics 39:49 — Q&A

Y Combinator

112,954 views • 1 month ago

Gemini 3, scaling laws and the 'finite data' era: my conversation with Sebastian Borgeaud, research engineer at Google DeepMind and a pre-training lead for Gemini 3 00:00 – Cold intro: “We’re ahead of schedule” + AI is now a system 00:58 – Oriol Vinyals's “secret recipe”: better pre- + post-training 02:09 – Why AI progress still isn’t slowing down 03:04 – Are models actually getting smarter? 04:36 – Two–three years out: what changes first? 06:34 – AI doing AI research: faster, not automated 07:45 – Frontier labs: same playbook or different bets? 10:19 – Post-transformers: will a disruption happen? 10:51 – DeepMind’s advantage: research × engineering × infra 12:26 – What a Gemini 3 pre-training lead actually does 13:59 – From Europe to Cambridge to DeepMind 18:06 – Why he left RL for real-world data 20:05 – From Gopher to Chinchilla to RETRO (and why it matters) 20:28 – “Research taste”: integrate or slow everyone down 23:00 – Fixes vs moonshots: how they balance the pipeline 24:37 – Research vs product pressure (and org structure) 26:24 – Gemini 3 under the hood: MoE in plain English 28:30 – Native multimodality: the hidden costs 30:03 – Scaling laws aren’t dead (but scale isn’t everything) 33:07 – Synthetic data: powerful, dangerous? 35:00 – Reasoning traces: what he can’t say (and why) 37:18 – Long context + attention: what’s next 38:40 – Retrieval vs RAG vs long context 41:49 – The real boss fight: evals (and contamination) 42:28 – Alignment: pre-training vs post-training 43:32 – Deep Think + agents + “vibe coding” 46:34 – Continual learning: updating models over time 49:35 – Advice for researchers + founders 53:35 – “No end in sight” for progress + closing

Gemini 3, scaling laws and the 'finite data' era: my conversation with Sebastian Borgeaud, research engineer at Google DeepMind and a pre-training lead for Gemini 3 00:00 – Cold intro: “We’re ahead of schedule” + AI is now a system 00:58 – Oriol Vinyals's “secret recipe”: better pre- + post-training 02:09 – Why AI progress still isn’t slowing down 03:04 – Are models actually getting smarter? 04:36 – Two–three years out: what changes first? 06:34 – AI doing AI research: faster, not automated 07:45 – Frontier labs: same playbook or different bets? 10:19 – Post-transformers: will a disruption happen? 10:51 – DeepMind’s advantage: research × engineering × infra 12:26 – What a Gemini 3 pre-training lead actually does 13:59 – From Europe to Cambridge to DeepMind 18:06 – Why he left RL for real-world data 20:05 – From Gopher to Chinchilla to RETRO (and why it matters) 20:28 – “Research taste”: integrate or slow everyone down 23:00 – Fixes vs moonshots: how they balance the pipeline 24:37 – Research vs product pressure (and org structure) 26:24 – Gemini 3 under the hood: MoE in plain English 28:30 – Native multimodality: the hidden costs 30:03 – Scaling laws aren’t dead (but scale isn’t everything) 33:07 – Synthetic data: powerful, dangerous? 35:00 – Reasoning traces: what he can’t say (and why) 37:18 – Long context + attention: what’s next 38:40 – Retrieval vs RAG vs long context 41:49 – The real boss fight: evals (and contamination) 42:28 – Alignment: pre-training vs post-training 43:32 – Deep Think + agents + “vibe coding” 46:34 – Continual learning: updating models over time 49:35 – Advice for researchers + founders 53:35 – “No end in sight” for progress + closing

Matt Turck

51,484 views • 9 months ago

Intelligence and Experience are orthogonal vectors Terence Tao is perhaps the world’s smartest person, but drop him into an accounting firm or onto a construction site and on day one he’s not going to be very productive Trajectory calls this The Experience Gap, and they have a way to close it Arjun Karanam explained how at our Sovereign AI event: 00:00 Introduction 00:12 Building the platform for continual learning 01:33 The experience gap: models have IQ but no tenure 02:52 Traceability → model spec → better models and harnesses 05:27 Four wishes for the agent ecosystem 06:34 Wish 1: Trace the whole tree — and capture the corrections 08:03 Wish 2: Evals from real traffic, graded in the real harness 09:26 Wish 3: Let the agents cook, and make tool responses informative 10:34 Wish 4: Get comfortable on open weights, experiment with routers 11:51 Why owning your intelligence shouldn't be consulted away 13:15 Demo: import a benchmark, train a model, deploy it 14:28 Q&A: What's the trainable object — weights, harness, or context? 16:08 Q&A: Continual learning without training on customer data 17:13 Q&A: Episodic memory and the hierarchy of feedback 19:37 Q&A: Where continual learning matters most

Intelligence and Experience are orthogonal vectors Terence Tao is perhaps the world’s smartest person, but drop him into an accounting firm or onto a construction site and on day one he’s not going to be very productive Trajectory calls this The Experience Gap, and they have a way to close it Arjun Karanam explained how at our Sovereign AI event: 00:00 Introduction 00:12 Building the platform for continual learning 01:33 The experience gap: models have IQ but no tenure 02:52 Traceability → model spec → better models and harnesses 05:27 Four wishes for the agent ecosystem 06:34 Wish 1: Trace the whole tree — and capture the corrections 08:03 Wish 2: Evals from real traffic, graded in the real harness 09:26 Wish 3: Let the agents cook, and make tool responses informative 10:34 Wish 4: Get comfortable on open weights, experiment with routers 11:51 Why owning your intelligence shouldn't be consulted away 13:15 Demo: import a benchmark, train a model, deploy it 14:28 Q&A: What's the trainable object — weights, harness, or context? 16:08 Q&A: Continual learning without training on customer data 17:13 Q&A: Episodic memory and the hierarchy of feedback 19:37 Q&A: Where continual learning matters most

Pat Grady

77,295 views • 1 month ago

2025 #지훈 챌린지 & 릴스 모음💝 00:00 sticky 00:16 샹하이 로맨스 00:40 Boom Boom Base (remix ver.) 01:01 오늘만 I LOVE YOU 01:30 💥💥💥 01:44 Make A Wish 02:20 뱅(Bang) ! 02:39 Kiss Kiss Shy Shy 02:52 wall🪼❤️‍🔥 03:07 🍉🍓🍊👌 03:24 Sweet Dreams 03:52 Lucifer 04:21 WE LIKE 2 PARTY 04:51 かわいいだけじゃだめですか？ 05:08 Ayo 05:36 BEBE 05:56 NIN 06:10 Come on!⚡️ 06:24 GO 지훈🎉 GO 지훈🥳 06:40 🪼💸 07:07 BANDIT 07:32 KNOW ABOUT ME 07:53 뭘봐 08:15 Cool & hot 08:38 SOURPATCH 08:55 Chains 09:24 이뻐이뻐 09:55 행복 10:22 move like that🎵 10:40 Shake It To The Max✌️ 10:52 boy for the 42🩵 11:08 JIHOON Calling📞 11:13 I Feel Good 11:34 HANDS UP 12:12 ROCKSTAR 12:36 Thunder 12:59 Bad Desire 13:17 시끄러!! 13:26 Touch It🪼 13:43 good vibes☀️ 13:52 Killin' It Girl 14:16 ECHO! 14:41 🚘🕶️ 14:56 Buzzin💭 15:07 FAMOUS 15:18 빌려온 고양이 15:35 Domino 15:51 The Blue Sun 16:07 POP 16:32 シュガリードライブ 16:52 훈제란즈✌️✌️ 17:02 되고파 너의 17:17 🪼vs🪼 17:34 before sleep 🛌 17:44 Walk Like👟 17:54 🌼 18:02 Barbie Girl 18:11 HOWL 18:23 Tempo 18:43 🪼🔄 18:52 Dougie🤙 19:10 What You Want 19:37 😎😎 19:51 Bang🪼 19:58 ✌️✌️ 20:10 EZ 20:29 M.O. 20:55 I‘m right here 💐 21:12 산책 21:24 body 21:42 SPAGHETTI 22:15 Blue Valentine 22:50 Hollywood Action 23:17 BURNING UP 23:40 하얀 그리움 24:20 Break your legs👟 24:29 Let‘s 🪩 24:47 Talk to You 25:16 My Body 25:37 MISMATCH 26:02 Do you wanna dance?🎵 26:12 🦾 26:41 All I Want for Christmas Is You 26:58 Hmm🕶️ 27:08 ˏ₍⸜̠̇⸝̠̇₎ˎ ̀⁽⸌̠̇⸍̠̇⁾ ́ˏ₍⸜̠̇⸝̠̇₎ˎ ̀⁽⸌̠̇⸍̠̇⁾ ́ 27:22 You like it? 27:28 락 (樂) 27:48 Love Shot

2025 #지훈 챌린지 & 릴스 모음💝 00:00 sticky 00:16 샹하이 로맨스 00:40 Boom Boom Base (remix ver.) 01:01 오늘만 I LOVE YOU 01:30 💥💥💥 01:44 Make A Wish 02:20 뱅(Bang) ! 02:39 Kiss Kiss Shy Shy 02:52 wall🪼❤️‍🔥 03:07 🍉🍓🍊👌 03:24 Sweet Dreams 03:52 Lucifer 04:21 WE LIKE 2 PARTY 04:51 かわいいだけじゃだめですか？ 05:08 Ayo 05:36 BEBE 05:56 NIN 06:10 Come on!⚡️ 06:24 GO 지훈🎉 GO 지훈🥳 06:40 🪼💸 07:07 BANDIT 07:32 KNOW ABOUT ME 07:53 뭘봐 08:15 Cool & hot 08:38 SOURPATCH 08:55 Chains 09:24 이뻐이뻐 09:55 행복 10:22 move like that🎵 10:40 Shake It To The Max✌️ 10:52 boy for the 42🩵 11:08 JIHOON Calling📞 11:13 I Feel Good 11:34 HANDS UP 12:12 ROCKSTAR 12:36 Thunder 12:59 Bad Desire 13:17 시끄러!! 13:26 Touch It🪼 13:43 good vibes☀️ 13:52 Killin' It Girl 14:16 ECHO! 14:41 🚘🕶️ 14:56 Buzzin💭 15:07 FAMOUS 15:18 빌려온 고양이 15:35 Domino 15:51 The Blue Sun 16:07 POP 16:32 シュガリードライブ 16:52 훈제란즈✌️✌️ 17:02 되고파 너의 17:17 🪼vs🪼 17:34 before sleep 🛌 17:44 Walk Like👟 17:54 🌼 18:02 Barbie Girl 18:11 HOWL 18:23 Tempo 18:43 🪼🔄 18:52 Dougie🤙 19:10 What You Want 19:37 😎😎 19:51 Bang🪼 19:58 ✌️✌️ 20:10 EZ 20:29 M.O. 20:55 I‘m right here 💐 21:12 산책 21:24 body 21:42 SPAGHETTI 22:15 Blue Valentine 22:50 Hollywood Action 23:17 BURNING UP 23:40 하얀 그리움 24:20 Break your legs👟 24:29 Let‘s 🪩 24:47 Talk to You 25:16 My Body 25:37 MISMATCH 26:02 Do you wanna dance?🎵 26:12 🦾 26:41 All I Want for Christmas Is You 26:58 Hmm🕶️ 27:08 ˏ₍⸜̠̇⸝̠̇₎ˎ ̀⁽⸌̠̇⸍̠̇⁾ ́ˏ₍⸜̠̇⸝̠̇₎ˎ ̀⁽⸌̠̇⸍̠̇⁾ ́ 27:22 You like it? 27:28 락 (樂) 27:48 Love Shot

🏝️

106,001 views • 9 months ago