Follow-up: why vertical AI timelines won't happen, per the weight-inspectability argument
ryunuck · x · 2026-09-21
Extending his argument that deep learning cannot inspect weights to locate inaccuracies, the author contends vertical timelines won't happen: even if you could target holes in the model's epistemics, fixes would make other learnings more useful or reshape the topology of holes — inaccuracy just shifts elsewhere and compounds exponentially.
Related event: Researcher: RL Agents Should Pick Tasks; Deep Learning Faces Limits(3 posts)→
More from AGI Musings
- Researchers ask: why is there no AI-first scientific journal with AI-run peer review? — anshulkundaje · 2026-09-21
- iamtrask: AI could reverse the centralization of art away from local creators — iamtrask · 2026-09-21
- Waypoint Bio argues AI in biology has a data problem before a model problem — anshulkundaje · 2026-09-21
- mmitchell_ai pushes back on 'stochastic parrot' critique as a category error — mmitchell_ai · 2026-09-21
- New paper estimates LLMs raised US natural unemployment rate by 0.1-0.2 points — mattbeane · 2026-09-21
- iamtrask says AI is heading the wrong way, pushes a pro-artist direction — iamtrask · 2026-09-21