AI Drug Discovery Expert Questions: Why Would RSI Differ from Traditional Optimizers?
rbhar90 · x · 2026-08-13
AI drug discovery expert rbhar90 questions the uniqueness of recursive self-improvement (RSI). He notes that automatic tuners and optimizers (e.g., Optuna, Vizier) have existed for long but haven't led to superintelligence. Research progress is often multifactorial; speeding one part just makes the next the bottleneck. Drawing from his experience, AI-driven drug discovery works decently for hit finding, but everything after remains hard, and drug discovery hasn't sped up significantly. He asks why RSI would be different.
More from AGI Musings
- Seroter's Daily Reading List: Why local models won't win, TDD in agent loop, and more — rseroter · 2026-08-13
- Debate: Agent skills may be obsolete as AGI should understand vague instructions — flowersslop · 2026-08-13
- AGI Progress Outpaces Forecasts: 19 Out of 24 2026 Predictions Already True — zetalyrae · 2026-08-13
- Eve of AGI: Continual Learning and Compaction to Disrupt Fine-Tuning — akbirthko · 2026-08-13
- Fermi Paradox as AGI Defense: No Alien AI Means No Galactic Disaster — francoisfleuret · 2026-08-13
- Will Small Language Models Catch Up to Frontier LLMs Soon? — elie2222 · 2026-08-13