Redditor defines AGI/RSI/ASI concretely, claims GPT-6 'Astra' already crossed the line
Crayola_God · reddit · 2026-10-16
Tired of undefined AGI arguments, a Redditor proposes concrete definitions: AGI = succeeding at tasks it wasn't trained for, with ARC-AGI-3 as the single deciding benchmark; RSI = self-improvement without human input (noting early RSI may be slow/costly before any 'foom'); ASI = outperforming every human on every task. They claim 'GPT-6 Astra' scored 99.9% on ARC-AGI-3, plays Minecraft, beats Portal, and operates robots — concluding AGI is achieved and RSI is near. Note: 'GPT-6 Astra' is not a verified real model; treat this as speculative discussion.
More from AGI Musings
- Sabine Hossenfelder fires back: academia's impact metric turned physics into a paper mill — skdh · 2026-10-16
- Researcher credits GPT for multiple accepted papers, incl. disproving Wegner's 1965 conjecture — ctjlewis · 2026-10-16
- dlwh: We optimized the legible version of our goals — and forgot the computation we meant to speed up — dlwh · 2026-10-16
- Ideas no longer have a waiting period: agents can build them as soon as they're articulable — akbirthko · 2026-10-16
- Ex-researcher: labs willing to cheat/distill ship great models faster than those building self-sufficient stacks — _aidan_clark_ · 2026-10-16
- Terry Tao reportedly still expects Riemann Hypothesis proof to be human+AI effort around 2050 — basedjensen · 2026-10-16