Forecasting community: stop using poorly-defined 'AGI' — define before you measure
davidmanheim · x · 2026-09-16
AI forecasting researcher David Manheim reiterated his 2019-2020 argument that the forecasting community should stop using poorly-defined terms like 'AGI', citing a self-deprecating contrast: forecasters say 'stop using vague terms', the AI community says 'we're building AGI!', and everyone else asks 'so is this new thing AGI?'
Answering how one would measure AGI, he stressed the key trick is to define terms before measuring, or you end up with disputed terminology and largely meaningless measurements — a pointed observation amid labs self-declaring AGI by their own internal standards.
Related event: AI Forecaster Urges Dropping the Ill-Defined Term AGI(2 posts)→
More from AGI Musings
- Fine-tuned small model matches 1T-param frontier model, a bull case for Periodic Labs — xeophon · 2026-09-16
- Economist: best path through AI-driven change is retiring older workers, hiring young into new jobs — paulnovosad · 2026-09-16
- Zvi mocks how professing high p(doom) became a status signal among researchers — TheZvi · 2026-09-16
- DeepSeek Engineer: I Optimize the Operators That Will Eventually Replace Me — minzhu0305 · 2026-09-16
- AI Raises Average Output Quality but Collapses Variance, Chopra Argues — paraschopra · 2026-09-16
- Economist: Don't Overread 'Negative Wage Effects of AI' Papers — paulnovosad · 2026-09-16