Michael Bronstein: genuine AI-driven scientific breakthroughs are still years away — scaling laws don't apply
mmbronstein · x · 2026-09-24
Asked for a timeframe on genuine groundbreaking scientific discovery by AI, Oxford professor and graph deep learning pioneer Michael Bronstein said his gut feeling is "a few years away." He explained that scaling laws do not apply here, since discovery typically involves real-world experiments that compute or algorithmic scaling cannot directly extrapolate.
More from AGI Musings
- Stewart Alsop: AI's real value is who can pull aesthetics out of latent space, not hard math — StewartalsopIII · 2026-09-24
- Researchers urge honesty about AI progress after LLM's math shock — aran_nayebi · 2026-09-24
- Personal identity has no determinate answers, argues dioscuri — dioscuri · 2026-09-24
- The Parfitian debate on personal identity in the age of AI minds — dioscuri · 2026-09-24
- Gallup: 90% of AI-aware adults in China report positive feelings about AI, world-leading — Polymarket · 2026-09-24
- Are humans just the bootloader? A causal-loop thought experiment on intelligence — Busy_Requirement_246 · 2026-09-24