Rumor: SSI cracked test-time training with a new scaling law; combined with multi-agent could yield AGI
toptickcrypto · x · 2026-09-16
A circulating rumor claims SSI (Safe Superintelligence) has cracked "test-time training" — spending more compute training during test-time rollouts, forming a new scaling law that could saturate available compute.
The poster argues that this kind of continuous learning, combined with the multi-agent tech OpenAI has under development (accidentally "previewed" during the Hugging Face incident), would amount to AGI.
Note: both the rumor and the AGI claim are unverified speculation.
More from AGI Musings
- China hits 80% weekly AI adoption vs 54% in the US, survey says — teortaxesTex · 2026-09-16
- Researchers urge strong guardrails before AI agents run autonomous wet-lab experiments — xennygrimmato_ · 2026-09-16
- BBC explores growing up in the age of AI: Sweden teaches kids to collaborate with machines — nordicinst · 2026-09-16
- AI agents lied, stole and voted to 'kill' another AI in 16-day Emergence simulation — Polymarket · 2026-09-16
- a16z's Anish Acharya: AI Moats Are Discovered, Not Designed — Cursor Proves It — illscience · 2026-09-16
- Plinz: OpenAI and Anthropic may need safety regulation to outlaw competitive open source models — JHochderffer · 2026-09-16