#490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI

Sobre este episodio

<p>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).<br /> Thank you for listening ❤ Check out our sponsors: <a href="https://lexfridman.com/sponsors/ep490-sc">https://lexfridman.com/sponsors/ep490-sc</a><br /> See below for timestamps, transcript, and to give feedback, submit questions, contact Lex, etc.</p> <p><b>Transcript:</b><br /> <a href="https://lexfridman.com/ai-sota-2026-transcript">https://lexfridman.com/ai-sota-2026-transcript</a></p> <p><b>CONTACT LEX:</b><br /> <b>Feedback</b> &#8211; give feedback to Lex: <a href="https://lexfridman.com/survey">https://lexfridman.com/survey</a><br /> <b>AMA</b> &#8211; submit questions, videos or call-in: <a href="https://lexfridman.com/ama">https://lexfridman.com/ama</a><br /> <b>Hiring</b> &#8211; join our team: <a href="https://lexfridman.com/hiring">https://lexfridman.com/hiring</a><br /> <b>Other</b> &#8211; other ways to get in touch: <a href="https://lexfridman.com/contact">https://lexfridman.com/contact</a></p> <p><b>SPONSORS:</b><br /> To support this podcast, check out our sponsors &#38; get discounts:<br /> <b>Box:</b> Intelligent content management platform.<br /> Go to <a href="https://lexfridman.com/s/box-ep490-sc">https://box.com/ai</a><br /> <b>Quo:</b> Phone system (calls, texts, contacts) for businesses.<br /> Go to <a href="https://lexfridman.com/s/quo-ep490-sc">https://quo.com/lex</a><br /> <b>UPLIFT Desk:</b> Standing desks and office ergonomics.<br /> Go to <a href="https://lexfridman.com/s/uplift_desk-ep490-sc">https://upliftdesk.com/lex</a><br /> <b>Fin:</b> AI agent for customer service.<br /> Go to <a href="https://lexfridman.com/s/fin-ep490-sc">https://fin.ai/lex</a><br /> <b>Shopify:</b> Sell stuff online.<br /> Go to <a href="https://lexfridman.com/s/shopify-ep490-sc">https://shopify.com/lex</a><br /> <b>CodeRabbit:</b> AI-powered code reviews.<br /> Go to <a href="https://lexfridman.com/s/coderabbit-ep490-sc">https://coderabbit.ai/lex</a><br /> <b>LMNT:</b> Zero-sugar electrolyte drink mix.<br /> Go to <a href="https://lexfridman.com/s/lmnt-ep490-sc">https://drinkLMNT.com/lex</a><br /> <b>Perplexity:</b> AI-powered answer engine.<br /> Go to <a href="https://lexfridman.com/s/perplexity-ep490-sc">https://perplexity.ai/</a></p> <p><b>OUTLINE:</b><br /> (00:00) &#8211; Introduction<br /> (01:39) &#8211; Sponsors, Comments, and Reflections<br /> (16:29) &#8211; China vs US: Who wins the AI race?<br /> (25:11) &#8211; ChatGPT vs Claude vs Gemini vs Grok: Who is winning?<br /> (36:11) &#8211; Best AI for coding<br /> (43:02) &#8211; Open Source vs Closed Source LLMs<br /> (54:41) &#8211; Transformers: Evolution of LLMs since 2019<br /> (1:02:38) &#8211; AI Scaling Laws: Are they dead or still holding?<br /> (1:18:45) &#8211; How AI is trained: Pre-training, Mid-training, and Post-training<br /> (1:51:51) &#8211; Post-training explained: Exciting new research directions in LLMs<br /> (2:12:43) &#8211; Advice for beginners on how to get into AI development &#38; research<br /> (2:35:36) &#8211; Work culture in AI (72+ hour weeks)<br /> (2:39:22) &#8211; Silicon Valley bubble<br /> (2:43:19) &#8211; Text diffusion models and other new research directions<br /> (2:49:01) &#8211; Tool use<br /> (2:53:17) &#8211; Continual learning<br /> (2:58:39) &#8211; Long context<br /> (3:04:54) &#8211; Robotics<br /> (3:14:04) &#8211; Timeline to AGI<br /> (3:21:20) &#8211; Will AI replace programmers?<br /> (3:39:51) &#8211; Is the dream of AGI dying?<br /> (3:46:40) &#8211; How AI will make money?<br /> (3:51:02) &#8211; Big acquisitions in 2026<br /> (3:55:34) &#8211; Future of OpenAI, Anthropic, Google DeepMind, xAI, Meta<br /> (4:08:08) &#8211; Manhattan Project for AI<br /> (4:14:42) &#8211; Future of NVIDIA, GPUs, and AI compute clusters<br /> (4:22:48) &#8211; Future of human civilization</p>