3 Comments
User's avatar
Tom Davidson's avatar

(inference costs on the METR tasks)

Tom Davidson's avatar

Great post .

Shouldn't it be easy to check this?

Just look at whether the doubling time for inference costs is faster than the doubling time for horizon length.

Anders Cairns Woodruff's avatar

If you're interested in this question, I have a more detailed empirical analysis here:

https://blog.redwoodresearch.org/p/ais-capability-improvements-havent