Capability gains

Ian Goodfellow says deep-learning capability growth will outlast short-term setbacks

Ian Goodfellow argued that disappointment with one model generation is a small speed bump within a multi-decade trend and forecast that deep-learning systems will continue growing in capability and importance. This records his dated forecast, not proof that the trajectory will continue.

0 comments · 0 votesOpen discussion

Public discussion is readable by everyone. Sign in to comment, reply, or vote.

No comments yet. Start the discussion.

CURRENT ASSESSMENT · REVISION 1
TOWARD DOOM28confidence 54/100

Why it moved the index

The forecast directly concerns continued growth in broadly transferable AI capability, a core pathway toward increasingly consequential agents. Magnitude is 28 because it is a general trajectory judgment rather than a demonstrated new capability or near-term breakpoint. Confidence is 54 because the first-person interview makes attribution certain and supplies a long historical comparison, while the forecast remains qualitative and its DoomBench effect requires one inference.

AUDIT TRAIL

Assessment history

  1. R1
    Toward 28 · confidence 54

    New historical first-person forecast not represented by Ian Goodfellow's existing labor evidence.

    17 Aug 2026
SHARE THE FINDINGS

Share this page

DoomBench social sharing card for Ian Goodfellow says deep-learning capability growth will outlast short-term setbacks.
  1. DoomBench assesses “Ian Goodfellow says deep-learning capability growth will outlast short-term setbacks” as evidence moving toward doom, with magnitude 28 and confidence 54 out of 100 in the capability gains category.

  2. The DoomBench assessment of “Ian Goodfellow says deep-learning capability growth will outlast short-term setbacks” is based on reporting from Le Random and records the editorial rationale, source quality, attribution, and revision history.

  3. DoomBench summarizes “Ian Goodfellow says deep-learning capability growth will outlast short-term setbacks” as follows: Ian Goodfellow argued that disappointment with one model generation is a small speed bump within a multi-decade...

    https://www.doombench.com/news/ian-goodfellow-says-deep-learning-capability-growth-will-outlast-short-term-setbacks-2025-11-04