Smash Notes

Lenny's Podcast: Product | Career | Growth

Inside the black box of AI: How AI models actually think | Emmanuel Ameisen (Anthropic)

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…

Transcript

People

Notes

  1. How Does an Artificial Intelligence Model Produce the Next Word?

    at 3:11 · Transcript

    Emmanuel Ameisen says a model takes a bunch of words, turns the words into numbers, does math on the numbers, and gets intermediate results. Full answer

  2. How Do Language Models Rhyme Without Saying the Plan?

    at 15:04 · Transcript

    Emmanuel Ameisen says a language model can write a coherent report or code even though it outputs only one word at a time. Full answer

  3. Why Does More Compute Keep Improving Word Prediction?

    at 26:46 · Transcript

    Emmanuel Ameisen says pouring more compute and more data into the task of predicting the next word makes the model's skill at predicting the next word keep… Full answer

  4. Why Is Alignment Still Hard If Watching Works?

    at 33:11 · Transcript

    Emmanuel Ameisen says that watching a model and catching it if it does bad stuff is, in one sense, totally the solution to alignment. Full answer

  5. Why Did an Anthropic Model Say a Demon Possessed It?

    at 36:50 · Transcript

    Emmanuel Ameisen says a model his team was training would go crazy on math problems. Full answer

  6. How Did Claude Beat Emmanuel Ameisen After Years?

    at 48:20 · Transcript

    Emmanuel Ameisen says an incredibly simple process, next-token prediction on text, is already breaking new ground in mathematics. Full answer

  7. Why Is Emmanuel Ameisen Optimistic About Model Alignment?

    at 51:59 · Transcript

    Emmanuel Ameisen is personally optimistic they will figure it out and keep models aligned with our interests. Full answer

  8. Why Do Artificial Intelligence Models Learn the Same Way?

    at 56:25 · Transcript

    Emmanuel Ameisen says the structure inside artificial intelligence models is surprisingly similar even when the models were trained years apart, on different… Full answer

  9. How Do Language Models Spot Bad Math Without Calculating?

    at 1:02:10 · Transcript

    Emmanuel Ameisen says language models answer a question one way and check it another way. They can do math, and they can say the color of a banana. Full answer

  10. What Funny Thoughts Fill Emmanuel Ameisen's Meetings?

    at 1:10:13 · Transcript

    Emmanuel Ameisen says that even in serious situations you should aim to have some levity. Full answer

Related notes

Related episodes

Plays from this note through the rest of the episode.

Note

Weekly highlights

One letter a week.

The notes worth hearing, with the minute they were said.