underscored

@underscored

3 clips · 1 follower

Follow
Tag:ai-researchClear

In a separate open-ended research project where the AI was tasked with proposing and testing hypotheses about an open problem in AI safety, Anthropic's models significantly outperformed two human researchers (97% performance improvement vs. 23%) when given a similar time budget (5 to 7 days).

Tim Fist
5d ago

For the ceiling, he showed the count can never exceed n to the power 1 plus 1 divided by 3, about n^1.333. For the floor, he exhibited explicit point sets where the count reaches at least n to the power 1 plus c divided by log log n, a formula that grows faster than n but only just barely. The chasm between these two bounds — stretching across eight decades and defeating every attempt to close it — is exactly what an AI system cracked open in 2025.

1mo ago

Underscored — save the words that stop you in your tracks.

Start saving quotes →