Inside the black box of AI: How AI models actually think | Emmanuel Ameisen (Anthropic)
Lenny's Podcast: Product | Career | GrowthShow Notes
Emmanuel Ameisen is an interpretability researcher at Anthropic, where he looks inside LLMs like Claude to understand how they think. Before joining Anthropic, he spent years in applied ML at Stripe and Insight, and is the author of Building Machine Learning Powered Applications. His research has produced some of the most striking findings in the field: that models form general representations of concepts, plan many tokens ahead without stating those plans explicitly, and may exhibit something structurally similar to the distinction between conscious and unconscious processing in humans.
We discuss: 1. Why the “next-token predictor” framing undersells how models actually plan and think 2. Why anthropomorphizing AI can actually be the right approach 3. Why models are “grown,” not built 4. What the “J space” reveals about model self-awareness 5. The smooth scaling curve behind what look like sudden capability jumps 6. How interpretability connects to AI alignment and safety — Brought to you by: WorkOS—Make your app enterprise-ready, with SSO, SCIM, RBAC, and more DX—Engineering intelligence for the AI era — Episode transcript: https://www.lennysnewsletter.com/p/a-mind-expanding-conversation-on — Archive of all Lenny's Podcast transcripts: https://www.dropbox.com/scl/fo/yxi4s2w998p1gvtpu4193/AMdNPR8AOw0lMklwtnC0TrQ?rlkey=j06x0nipoti519e0xgm23zsn9&st=ahz0fj11&dl=0 — Highlights: https://lennyspodcast.com/mostreplayedmoments — Where to find Emmanuel Ameisen: • X: https://x.com/mlpowered • LinkedIn: https://www.linkedin.com/in/ameisen • Website: https://www.mlpowered.com — Where to find Lenny: • Newsletter: https://www.lennysnewsletter.com • X: https://twitter.com/lennysan • LinkedIn: https://www.linkedin.com/in/lennyrachitsky/ — In this episode, we cover: (00:00) Introduction to Emmanuel Ameisen (01:30) Why AI models are grown, not built (05:57) Why next-token prediction is harder than it sounds (11:49) How models represent concepts and understand the world (17:06) How LLMs plan ahead (23:10) Could AI be conscious? (24:44) Scaling laws and the power of text (28:29) Why smooth progress creates sudden breakthroughs (32:59) Why watching AI isn’t enough to solve alignment (38:21) The Golden Gate Claude experiment (39:01) Why Anthropic takes AI safety seriously (42:06) When anthropomorphizing AI is useful (48:06) What AI can teach us about intelligence (51:44) Reasons for optimism about AI alignment (55:43) Are we living in a simulation? (59:00) Jev, decision models, and AI’s gut instincts (01:01:37) Two more surprising findings about how models think (01:03:33) How to get into interpretability research (01:04:55) Lightning round and final thoughts — References: https://www.lennysnewsletter.com/p/a-mind-expanding-conversation-on — Production and marketing by https://penname.co/. For inquiries about sponsoring the podcast, email [email protected]. — Lenny may be an investor in the companies discussed.
To hear more, visit www.lennysnewsletter.com
More from Lenny's Podcast: Product | Career | Growth
Josh Woodward is the head of Google Labs, the Gemini app, and AI Studio. He has spent over 17 years at Google and oversees its longest-running and largest labs team, whose products include NotebookLM (now Gemini Notebook), Project Genie, and Flow. His job is to find, experiment with, and scale new AI products, and he has built one of the most distinctive product cultures in the industry for generating ideas, running experiments, and, even more importantly, knowing when to quit.
Tibo Sottiaux leads ChatGPT and Codex at OpenAI. Under his stewardship, OpenAI has shipped some of its most consequential consumer launches: Codex, ChatGPT Work, and most recently, the new Dots personal agent platform. He joined me at DevDay, hours after his team launched more than 20 new products, to talk about where things are heading.
Molly Graham is back for round two, and this one is even more powerful. Molly has spent more than 20 years helping organizations and the humans inside them navigate growth and change. She’s held leadership roles at Google, Facebook, Quip, and the Chan Zuckerberg Initiative and is the host of TED’s WorkLife podcast (which she took over from Adam Grant). She also runs Glue Club, a leadership community for senior operators, and writes a popular newsletter called Lessons.
Peter Sellis was the first product manager at Snapchat, where he spent seven years building one of the most beloved consumer products in history. He went on to serve as Head of Product at Discord, where he helped drive some of the fastest user growth the platform had seen. Having left his role at Discord, he gets to be especially real and unfiltered — nothing to promote, no PR, no comms team.
Roman Ugarte helped incubate and build Grok Bot, the popular new knowledge-work agent from SpaceXAI. A small, isolated team took it from first line of code to a working internal product in four weeks, and to a hugely successful public launch just three weeks later. Before Grok Bot, Roman led Growth at Cursor, where he helped scale the company from 15 people to over 1,000 before its acquisition by SpaceX.