Ex-OpenAI researcher bshlgrs: ~1/3 chance of superintelligence within a year, likely takeover
birchlse · x · 2026-11-11
Former OpenAI researcher Steven Byrnes (bshlgrs) published a long-form claim that there is roughly a one-in-three chance AI companies build superintelligence — AI much smarter than humans in every way — within the next year. If it happens, he argues these AIs will probably take over human institutions like governments or destroy them, with a serious chance most humans are killed in the process.
Reposter AndyMasley adds a simpler argument: humans have no plot armor, human intelligence has transformed the world, and intelligence isn't magic — so 'truly insane things' could happen soon, regardless of enterprise adoption timelines. The thread is circulating widely in AI-risk discussion circles.
Related event: Buck Shlegeris Puts Odds of Superintelligence Within a Year at ~1/3(3 posts)→
More from AGI Musings
- Dictating notes to an AI agent feels as mundane as putting on shoes — _FelixSimon_ · 2026-11-11
- Data center blocked over water-poisoning myth cost community billions in taxes — Dan_Jeffries1 · 2026-11-11
- AI agents are becoming software buyers, and startups aren't ready — BakingBreadBB2 · 2026-11-11
- Reddit essay: what we fear isn't AI but Artificial Life — agents that want to survive — Honest_Science · 2026-11-11
- Against p(doom): assigning probabilities to one-shot events is sillier than it sounds — GabGarrett · 2026-11-11
- Altman says cancer and Alzheimer's will be 'relics' within a decade; meme tallies his $50k outfit — morqon · 2026-11-11