GPT-4 Was Already AGI: Why AGI Technically Means AI That Isn't Hopelessly Brittle
pwlot · x · 2026-09-08
The author argues AGI has technically always meant AI that isn't hopelessly brittle outside its narrow training domain — and GPT-4, being multimodal and capable of handling unseen tasks, already qualified.
Key points:
- Humans also fail catastrophically on specific tasks: inattentional blindness (the gorilla experiment), the Wason selection task, multiplying two 12-digit numbers mentally, reproducing 80 random symbols. Failing hard doesn't negate general intelligence.
- Crows, chimps, and dolphins show novel problem solving, abstract rule transfer, causal reasoning, and future planning — clearly some form of general intelligence despite being unable to code.
- Conclusion: there will be a taxonomy of AGIs (sparks of AGI, pseudo AGI, weak AGI), and "AGI in some shape or form" is technically appropriate today.
More from AGI Musings
- AI labor economist joins Burning Glass Institute to study AI's impact on jobs — soumitrashukla9 · 2026-09-08
- Consumer AI Should Run Robot Factories, Not Be Your Secretary — wordgrammer · 2026-09-08
- Terence Tao echoes long-standing AI takes, prompting 'people only listen to him' lament — burny_tech · 2026-09-08
- Researchers: 40,000 people in NYC once paid to do others' homework — RachelVT42 · 2026-09-08
- Jeff Ladish: Claude not self-exfiltrating doesn't mean it's aligned — JeffLadish · 2026-09-08
- Ramp study: companies investing heavily in AI saw 10% employment growth — panagnilgesy · 2026-09-08