Lenny's Podcast Video posted an update
How models think | Emmanuel Ameisen (Anthropic)
Why it mattersThe description presents Emmanuel Ameisen's team's work decoding model activations to examine how a plan may form before output. A poetry example illustrates the claim: the model identifies a rhyme scheme internally, and later words reflect that decision.
Discuss: What does the poetry example suggest about the role of model activations in shaping the words a language model produces?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.