Tim Hua
banner
timhua.bsky.social
Tim Hua
@timhua.bsky.social
Helping people is good I guess

Trying to do AI interp and control

Used to do economics

timhua.me
a lot of the criticism about the methodology/statistics reads like "oh wow this macro paper used a panel data of only eight countries, you can't draw any conclusions from N = 8!!!" like yes you can. Skill issue on the part of the reader.
July 18, 2025 at 3:02 AM
well idk a lot of research isn't that useful.

I think this is a good study executed as well as they could've, and it's probably the best study of its type/genre? Also they're operating in an insanely brutal environment, where any result like this is ~obsolete after like five months. I just think
July 18, 2025 at 3:00 AM
The main point is to reject the expert predictions null.

But also come on, there's nothing special about the zero point. Even if it's not a slowdown surely the study contained valuable takeaways.
July 16, 2025 at 6:37 PM
"In other words the existence of a free market creates the most efficient production possible"
February 7, 2025 at 5:35 AM
Wait isn't switching costs for foundation model providers near zero
January 27, 2025 at 8:14 PM
Man
December 20, 2024 at 8:43 AM
Dude at the FAR Labs office corner I work at, there’s like four people (including me) and three of us would be on split keyboards it’s crazy
November 29, 2024 at 9:30 PM
Banger
November 29, 2024 at 9:29 PM