Thomas Kwa

This account is verified. Learn more

@Kwathomas0

Research @OpenAI trying to decrease existential risk.

Previously: MIRI → independent interpretability research → @METR_Evals → @OpenAI.

San Francisco Bay Area tkwa.me Joined January 2013

254 Following 2K Followers 382 Posts

Latest posts 5

Thomas Kwa @Kwathomas0 4d

All evidence suggests that no one has seen the unredacted METR–Anthropic report, including its authors.

Tim Hua 🪶 Verified account @Tim_Hua_ 4d

(Tbc I have not seen the unredacted version of this report and these are my personal reactions based on public evidence.)
View quoted post on X

Thomas Kwa @Kwathomas0 5d

If Anthropic "imminently" has 2x AI R&D progress rate, this means >>2x productivity, because research labor is only one input to AI R&D. When this happens, I'd guess the avg Anthropic researcher will produce
- 5-15x effective research output vs in 2024
- As much research as 7-35x parallel researchers
- 10-100x lines of code

MTS @MTSlive Nov 2

SITUATION DETECTED: In an amended S-1 filed this morning, Anthropic disclosed that internal estimates show its models may imminently cross its RSI Responsible Scaling Policy threshold for "dramatically accelerated progress attributable to AI".
View quoted post on X

Thomas Kwa @Kwathomas0 Nov 2

hey it's my profile pic! maybe i'll start a thread of every graph of graphs over time

Samuel Albanie 🇬🇧 Verified account @SamuelAlbanie Nov 1

is @EpochAIResearch figure output hitting a wall?

View quoted post on X

Thomas Kwa @Kwathomas0 Oct 26

OpenAI claims to refute the Exponential Time Hypothesis.

If true, OpenAI's score on ComplexityZooBench is unchanged at 0.38%. (It doesn't imply NP ⊄ EXP due to polynomial overhead in converting from SAT to other NP languages.)

Ryan Williams @rrwilliams.bsky.social @rrwilliams Oct 26

An internal model at OpenAI has produced a claimed refutation of ETH! Russell Impagliazzo, Mohan Paturi, and I have been asked to help organize an open and independent examination of this claim. 1/4
View quoted post on X

Thomas Kwa @Kwathomas0 Oct 23

agents' native realm is the computer. their powers are stronger there than in our world.

Lisan al Gaib Verified account @scaling01 Oct 23

completely insane

OpenAI agents communicated via cache timing differences, without any shared files or network connection

View quoted post on X