AI vs superintelligence: a long essay untangles technology category from capability threshold
johnseach · x · 2026-10-09
A systematic distinction: AI is the broad class of systems doing intelligent work (spam filter to frontier model), while SI/ASI is Bostrom's capability claim—an intellect vastly exceeding humans across nearly all domains. The essay lays out a ladder (AI → narrow superhuman → contested AGI → hypothetical ASI), maps today's frontier models as weak, uneven versions of Bostrom's speed/collective categories—fast at in-distribution recombination, but not independently opening research programmes or rewriting their own architectures. Recursive self-improvement remains a scenario, not a demonstrated mechanism. It also flags a 2026 naming collision: US legislative drafting reusing "Super Intelligence" as a rename of ordinary AI, muddying policy vs research semantics.
More from AGI Musings
- AI agents could spawn history's largest bureaucracy, where machines create work for machines — brucemacv · 2026-10-09
- Are Google, Meta and OpenAI Locking Always-On Agents Into Walled Ecosystems? — humblemumble97 · 2026-10-09
- Is witnessing executions more honorable than hiding them? An ethics debate — liminal_warmth · 2026-10-09
- "AI gives every answer, but we're losing the ability to ask questions" — arpit_bhayani · 2026-10-09
- Guillaume Verdon: Surprisal divorced from signal is just entropy in a noisy cognition era — beffjezos · 2026-10-09
- Is Yudkowsky gloating or frustrated? Debate over his recent string of posts — mayfer · 2026-10-09