Andrej Karpathy (安德烈·卡帕西)

(0:00) Gavin Baker joins the show! (0:30) Andrej Karpathy joins Anthropic; hypergrowth and profitability (12:42) Why Americans have turned on AI, anti-human perception (27:22) Trump pulls AI EO, US-China AI relationship, dystopian AI layoffs (45:19) SpaceX S-1 tear down! Breaking down the three major businesses and the case for a $2T valuation (1:11:22) Nvidia smashes earnings but stock falls, why

What if your AI's memory was mathematically verifiable? What if every retrieval was provenance-backed, every result bit-exact and cryptographically reproducible? OnMemory.ai introduces deterministic semantic memory built on E8 lattice quantization, replacing probabilistic vector search with a multi-lane retrieval engine where every answer can be traced to its source. In this session, Andrew K. Dav

Sophia Dew and Binji Pande speak with Vitalik Buterin about technology, human agency, and how the internet is changing the way people think, build, and relate to the world around them. Drawing from his writings and personal reflections, Buterin discusses how his worldview has evolved over the last decade, from creating Ethereum as a teenager to thinking more deeply about the social and philosophic

Eric Jang walks through how to build AlphaGo from scratch, but with modern AI tools. Sometimes you understand the future better by stepping backward. AlphaGo is still the cleanest worked example of the primitives of intelligence: search, learning from experience, and self-play. You have to go back to 2017 to get insight into how the more general AIs of the future might learn. Once he explained how

The future of automating knowledge work depends on AI agents that can reliably read and understand documents — but today's agents struggle with complex layouts, tables, and visual elements. This talk by LlamaIndex' Jerry Liu explores why document parsing remains a critical bottleneck for agentic workflows and introduces new open-source innovations to address it, including ParseBench, a benchmark f

Clem Delangue joins MTS to discuss the global open-source AI landscape, the current large language model bubble, and the future of consumer robotics. Originally aired on MTS, Theo Jaffee and Sofia Puccini speak with Clément Delangue, CEO at Hugging Face, about the global open-source AI race, why he believes the real bubble is in API-based large language models, and how robotics could become the ne

We are going to talk about agent memory. Um, I don't know about you, but after building some of these systems and and working with uh memory and context engineering, I kind of wish I could implement a system like this for myself. Um, my memory works just well enough that I can remember your name for about 30 seconds after I meet you. And then I just wish I could run a query and pull that out of my

I'm really excited to talk to you guys about what we're doing with the Reddus context engine and what my team, the applied AI team at Reddus, uh, has been building and how we built an S sur agent that you can actually trust in production. And as you guys might know, that's a feat. Let's start with the problem statement. So, why did my team build out an S sur agent in the first place? Um, when we w

Good afternoon. Welcome to OpenAI Forum. Uh my name is Chris Nicholson. I'm with the global affairs team and I'm glad to be here with all of you. So the forum, as some of you know, is a place where we talk with experts about how AI is being used in the world. Today's conversation is about codecs and why it matters beyond software engineering. So more and more people are using codeex to help with k

Anjney Midha, founder of AMP PBC, speaks with Ben Horowitz, cofounder of a16z, about how venture capital changed from a small, relationship-driven business into a scalable system for backing new technology companies. They discuss network effects, firm design, leadership, culture, and how AI is reshaping both the capital race and the kinds of companies that can be built now. Resources: Follow Ben o

Matt Bornstein speaks with Scott Chacon, cofounder of GitHub and CEO of GitButler, about why Git's user interface has barely changed since 2005, how GitButler is rethinking version control for both humans and AI agents, and what the "next GitHub" might actually look like. They cover parallel branches, agent-optimized CLI design, the future of code review, and why the best engineers of the future w

This episode originally aired on the Latent Space Podcast. swyx and Alessio Fanelli speak with Marc Andreessen about the arc of AI from its origins in 1943 to today's breakthroughs in reasoning, coding agents, and self-improvement. They cover the parallels between AI scaling laws and Moore's Law, the architectural insight behind Claude Code and the Unix shell, the coming supply crunch in compute,

(0:00) Jensen Huang joins the show! (0:26) Acquiring Groq and the inference explosion (8:53) Decision making at the world's most valuable company (10:47) Physical AI's $50T market, OpenClaw's future, the new operating system for modern AI computing (16:38) AI's PR crisis, refuting doomer narratives, Anthropic's comms mistakes (20:48) Revenue capacity, token allocation for employees, Karpathy's aut

David Haber speaks with Owen Jennings, executive officer and business lead at Block, about how the company rebuilt itself around AI agents, small squads, and internal tools like Goose and Builder Bot after restructuring more than 40% of its workforce. They discuss what it took to execute a major restructuring, how teams of three are now doing what teams of 14 used to, and how Block is shipping AI-

Jeff Kaplan is a legendary Blizzard game designer of World of Warcraft and Overwatch, now preparing to launch a new game, The Legend of California, from his new studio Kintsugiyama – available to wishlist on Steam today, with alpha later in March. Thank you for listening ❤ Check out our sponsors: https://lexfridman.com/sponsors/ep493-sc See below for timestamps, and to give feedback, submit questi

Nathan Lambert and Sebastian Raschka are machine learning researchers, engineers, and educators. Nathan is the post-training lead at the Allen Institute for AI (Ai2) and the author of The RLHF Book. Sebastian Raschka is the author of Build a Large Language Model (From Scratch) and Build a Reasoning Model (From Scratch). Thank you for listening ❤ Check out our sponsors: https://lexfridman.com/spons

I was recently chatting with a friend who has a similar job to mine. We were talking about how even though our jobs are fundamentally about learning about stuff, our time so easily gets sucked up by other things. So to hold myself accountable, I’m gonna try to publish a blog post every two weeks or so where I explain what I’ve been reading. Max Hodak’s theory of consciousness I’m totally gonna but

Peter Steinberger is the creator of OpenClaw, an open-source AI agent framework that’s the fastest-growing project in GitHub history. Thank you for listening ❤ Check out our sponsors: https://lexfridman.com/sponsors/ep491-sc See below for timestamps, transcript, and to give feedback, submit questions, contact Lex, etc. Transcript: https://lexfridman.com/peter-steinberger-transcript CONTACT LEX: Fe

My main bottleneck is finding excellent guests. So, I’m hiring a couple part time scouts to help me find the next David Reich/Sarah Paine/Adam Brown. $100/hour, fully remote, work hours are flexible - I expect it’ll be 5-10 hours a week. Ideal candidate is maybe a grad student, or a post doc, or working in one of the fields I wanna find guests in. I’m looking for people who are really plugged into

What are we scaling? I’m confused why some people have short timelines and at the same time are bullish on the current scale up of reinforcement learning atop LLMs. If we’re actually close to a human-like learner, this whole approach of training on verifiable outcomes is doomed. Currently the labs are trying to bake in a bunch of skills into these models through “mid-training” - there’s an entire

The mission I originally titled my podcast The Lunar Society. I changed it to Dwarkesh Podcast eventually because people kept thinking it was a crypto podcast (”to the moon!!!”). I named it after The Lunar Society of Birmingham, an informal club that met in the late 18th century. Members included James Watt, Matthew Boulton, Erasmus Darwin, Joseph Priestley, and Josiah Wedgwood. These were the sci

What are we scaling? I’m confused why some people have short timelines and at the same time are bullish on the current scale up of reinforcement learning atop LLMs. If we’re actually close to a human-like learner, this whole approach of training on verifiable outcomes is doomed. Currently the labs are trying to bake in a bunch of skills into these models through “mid-training” - there’s an entire

The Andrej Karpathy episode. During this interview, Andrej explains why reinforcement learning is terrible (but everything else is much worse), why AGI will just blend into the previous ~2.5 centuries of 2% GDP growth, why self driving took so long to crack, and what he sees as the future of education. It was a pleasure chatting with him. Watch on YouTube ; read the transcript . Sponsors * Labelbo

Richard Sutton is the father of reinforcement learning, winner of the 2024 Turing Award, and author of The Bitter Lesson. And he thinks LLMs are a dead end. After interviewing him, my steel man of Richard’s position is this: LLMs aren’t capable of learning on-the-job, so no matter how much we scale, we’ll need some new architecture to enable continual learning. And once we have it, we won’t need a

Sergey Levine , one of the world’s top robotics researchers and co-founder of Physical Intelligence , thinks we’re on the cusp of a “self-improvement flywheel” for general-purpose robots. His median estimate for when robots will be able to run households entirely autonomously? 2030. If Sergey’s right, the world 5 years from now will be an insanely different place than it is today. This conversatio

Trained for ~8000 episodes, each episode = ~30 games. Updates were done in batches of 10 episodes, so ~800 updates total. Policy network is a 2-layer neural net connected to raw pixels, with 200 hidden units. Trained with RMSProp and learning rate 1e-4. The final agent does not beat the hard-coded AI consistently, but holds its own. Should be trained longer, with ConvNets, and on GPU. This is ATAR

we all know and love archive it is a massive pre-print repository with a lot of wonderful papers unfortunately these papers are not very easy to skim search or sort new papers are also added every single day but unfortunately all you get is this huge featureless list it's easy to miss interesting papers that might be relevant to your research introducing archive sanity the function of this website

I'd like to point out that what I'll be presenting today is partly my work in collaboration with others and sometimes I'm presenting work done by people in my group uh that I wasn't really involved in but it's joint work with many many people you'll see lots of names throughout the talk so uh take that with a grain of salt uh so what I'm going to tell you about is kind of how Google got to where i

terms of administrative items um everyone should be done with sment 3 now if you're not done I think you're late and you're in trouble uh Milestone grades will be out very soon we're still getting through them and they're basically I think done but we have to double check a few things and we'll send them out okay so in terms of reminding you where we are in the class uh last class we looked very b

so our administrative uh points for today U assignment three is due tonight so who's done that's good uh was it easier than assignment two okay that's good um hopefully gives you more time to work on your projects um so also remember your Milestones were were uh turned in were turned in last week so we're in the process of looking through the Milestones to make sure those are okay and also we're w