Elektrine lite

← Feed

@lurker@awful.systems

2026-11-11 02:31 UTC

The AI 2027 authors are back: Q3.5 Timelines Update: Looks Like the Intelligence Explosion Will Happen in 2027 After All It’s all about that new METR data. Took a quick scroll and the whole thing is about forecast modelling. Its pretty lengthy so if anyones willing to take the plunge and read it all youll have my thanks METR recently estimated that Anthropic is now making AI progress about twice as fast as its historical rate (they would not be surprised by anything from 1.3x to 3x), and guessed that “perhaps ≈80%” of the current acceleration is due to AI. If they’re right, that implies an AI-driven acceleration that is 5 months ahead of the AI 2027 scenario. Going down a little bit: A 1.7x pace of progress is 5 months ahead of the AI 2027 scenario. AI 2027 predicted 2x AI R&D uplift in Dec 2026; as argued above, this would imply a substantially lower pace of progress than the 1.7x AI-attributable pace of progress that METR’s estimates imply for Nov. So when exactly did AI 2027 predict a 1.7x pace of progress over baseline? We created this trajectory by modifying a few key parameters until it roughly matched AI 2027. In that trajectory, 1.7x acceleration over the 2024.0–2025.5 rate of progress happens in roughly April 2027. So by this metric, we are ~5 months ahead of schedule. And a chart of their predictions to end it all off: Mark your calendars everybody!

Replies (6)

  • @lurker@awful.systems 2026-11-11 02:41

    So I went all the way down to the appendix, and found a gem: Appendix: Ambiguity in “80% of current acceleration is due to AI”. Claude Opus 5.5 gives the following possible interpretations of METR’s claim that “we guess that a majority (perhaps ≈80%) of the current (October 2026) acceleration is due to AI.” Claude Opus 5.5 gives the follow possible interpretations Did…did they just say they used Claude for this??

    Open ##5517335

  • @swlabr@awful.systems 2026-11-11 04:31

    TED-AI as in the meaninglessly distilled and contrived ted-talk idea of AI

    Open ##5517583

  • @Architeuthis@awful.systems 2026-11-11 07:02

    It’s all about that new METR data. Yeah, they seem to have moved on from multiplying sufficiently large matrixes is magic to thinking it’s actually the code writing that’ll get mystical any time now.

    Open ##5517983

  • @BioMan@awful.systems 2026-11-11 07:12

    I have had to explain to my boss that METR are cultists

    Open ##5518001

  • yeah but wen fuckable digital jesus.

    Open ##5519281

  • @TrashGoblin@awful.systems 2026-11-12 16:28

    Sufficiently overfitted benchmarks are indistinguishable from exponential progress.

    Open ##5524411