Why hasn't AI produced new scientific theories yet? One dev's capability-gap theory
voooooogel · x · 2026-10-18
Replying to discussion on why AI hasn't yet delivered the long-expected breakthroughs, voooooogel argues it comes down to missing capabilities: exactness plus the ability to filter out bad approaches without tools, the ability to iteratively refine broken solutions, and the ability to wander through idea space in chain-of-thought, assembling an effectively unbounded number of ideas.
He adds a broader musing: if future fields show a similar lag, we may later say "it just took a bit longer — we needed X capability," and wonders what that X will be.
More from AGI Musings
- Gary Marcus flags prompt injection: AI agents are easy prey for phishing emails — GaryMarcus · 2026-10-18
- MPI professor: I can advise 6 PhDs but only 2-4 AI agent projects at once — Michael_J_Black · 2026-10-18
- "Barbarians inside the castle walls": AI circle's bleak metaphor gets a rueful yes — MickeySteamboat · 2026-10-18
- Automating Experiments Makes Premise Choice the Real Bottleneck in Recursive Self-Improvement — Gauri_the_great · 2026-10-18
- Reddit asks: When will the AI bubble pop, and will it be a crash or slow deflation? — Alex__007 · 2026-10-18
- Blogger imagines ASI curing aging: your oldest friend could be someone you've known for 600 years — Dr_Singularity · 2026-10-18