19 - Mechanistic Interpretability with Neel Nanda
AXRP - the AI X-risk Research Podcast
00:00
Scaling Laws Are Less Useful for AI Expert Reduction or AI Alignment
I would predict that scaling laws were knowably much more useful than mechanistic interpretability. I actually think that the most important consequence was just this idea that scaling will continue to work and that we should just try really hard to make models bigger. It's seeming less and less likely there's some magic point where everything breaks and you've wasted a billion dollars because their ability to forecast is so tied to it. If you actually got good at mek and tup we'd be good at forecasting, that's probably going to call it mek and Tup.
Transcript
Play full episode
