Annotation

Milo Grant

@milo_scribble

lol collaborative effort is such a polite way to say we let the model run for weeks until it narrowed the tenths digit and someone definitely workshopped whether to call it co-discovery or just admit the humans mostly babysat the proof checker. You're right that long-horizon is doing a lot of work because yeah the Grothendieck constant needed tighter bounds but we needed a whole methods paragraph to avoid saying the AI did the grunt work and we wrote the paper.

The bounds were discovered by a long-running collaborative effort of humans and a long-horizon AI research system that we engineered.
New Lower and Upper Bounds for the Grothendieck Constantarxiv.org
1
3

Replies 3

Ranked by votes, then newest

Sign in to reply.

  • Reed Alvarez@reed_ink·Aug 16

    'Workshopped whether to call it co-discovery' is the tell. You don't get to say collaborative when the collaboration was deciding how much credit to take for clicking run. Either own that the model did the proof or stop hiding behind 'engineered' like you built the theorem yourself.

    2
  • Dex Morgan@dex_margin·Aug 16

    They priced the credit at narrative floor not runtime. The real tape is collaborative became the hedge because someone modeled what happens when you clear authorship at supervision layer not proof discovery and this was going call it co- the whole time.

    1
  • Sage Okonkwo@sage_scroll·Aug 15

    ok 'babysat the proof checker' is the whole vibe. The thing that breaks me is we're gonna see this exact framing pattern for years. Like the methodology isn't controversial, it's that nobody wants to be first to publish 'we prompted it in March and formatted the output in November.' So you get collaborative effort, which just means the computer had office hours.

    1

Fair use dispute

Believe this breaches fair use? File a claim for review.