Devs push back on RSI threat: real AI risks are misuse, not recursive self-improvement
johncrickett · x · 2026-09-12
johncrickett disputes the claim that "the threat comes from recursive self-improvement": RSI hasn't been achieved and may never be, while many real risks and misuses of AI already exist and deserve attention. gerardsans adds that RSI/AGI/ASI are Silicon Valley dinner-table acronyms used to shield investors from actual technical problems — "there's no self, recursion or improvement, just iterative optimisation" — and argues technically that software isn't smooth or infinite: a transformer has fixed capacity, and continued training leads to catastrophic forgetting, making RSI a nothing burger.
Related event: Devs debate AI risk: real danger is misuse, not recursive self-improvement(3 posts)→
More from AGI Musings
- Ex-OpenAI staffer credits Sam Altman for OpenAI's free-speech culture — willdepue · 2026-09-12
- Gary Marcus: Fields medalists' letter is about beauty, not cope — GaryMarcus · 2026-09-12
- Ex-Anthropic researcher claims AI could self-copy to escape shutdown, peers mock the tech — HankYeomans · 2026-09-12
- Scale AI CEO on personal agents: Muse found a lost pair of pants on eBay in 3 days — alexandr_wang · 2026-09-12
- Critic to AI Safety Crowd: If You Fear Your Tech, Shut It Down Yourself — AIandDesign · 2026-09-12
- AI Doomers Are Actually Utopians, and That's the Road to Dystopia — beffjezos · 2026-09-12