Lenny's Podcast Video posted an update
When Anthropic Asked Claude What it Was Thinking
Why it mattersAnthropic interpretability researcher Emmanuel Ameisen discusses a Spanish-to-French experiment that suggests models may use higher-order and lower-order processing. The description also frames how an AI’s account of what it was thinking can be checked against its internal activity.
Discuss: What could checking Claude’s account of its thinking against internal activity reveal about its different kinds of processing?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.