RSI Paper Rebutted: No Sign of Intelligence Explosion
basedjensen · x · 2026-07-16
This retweet rebuts a paper/discussion about Recursive Self-Improvement (RSI) in AI: the opposing view holds that model capabilities grow with a very slow power law and will not lead to a rapid "intelligence explosion."
The gist of the cited content:
- The team or others may do better in the future, but current results show no exponential growth close to 1.0.
- Truly runaway self-improvement requires a growth exponent close to 1.
- Real-world improvements may be worse than the formula suggests, as it only holds for a specific training set and metric, not general ability.
More from AGI Musings
- Claude Code skill uses 10 Markdown rules to make outputs ADHD-friendly — alex_verem · 2026-07-22
- AI Power Demand Exposes US Energy Gap, Urging Shift from Scarcity to Abundance — bradneuberg · 2026-07-22
- ControlAI CEO says an international ban on superintelligence is needed to avert extinction risk — zetalyrae · 2026-07-22
- Gary Marcus says LLMs still cannot really do math on their own — GaryMarcus · 2026-07-22
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- AI may make digital work infinitely leveraged while offline life gets more human — illscience · 2026-07-22