A practical AGI test: replace one human validating 150,000 document conversions
burkov · x · 2026-09-27
burkov (ChapterPal) proposes a very concrete AGI test. His fully automatic pipeline converts PDFs, arXiv preprints, and thousand-page books with thousands of tables, figures, and equations into near-perfect Markdown — and he has the cash to ingest 150,000 items overnight, building the world's largest knowledge collection optimized for AI-tutor self-learning. The bottleneck: 20% of converted items have minor issues (leaked LaTeX, wrong years, wrong PDF entirely), and no AI can reliably detect all issues, fix them, and publish without missing problems or breaking things that weren't broken. He is the sole bottleneck; the moment an AI can replace him as a pre-publishing validator/fixer, he'll declare AGI built. Both an empirical probe of agent limits and a personal AGI criterion.
More from AGI Musings
- Would Benevolent Machine Gods Tolerate Despotic Regimes? X Users Debate AI Theodicy — dioscuri · 2026-09-27
- Emily Bender's FAQ on LLM limits resurfaces, framing called 'extremely relevant' to current models — zacharynado · 2026-09-27
- 24 senior mathematicians at Harvard summit: PhDs should no longer hinge on dissertations — alejandroll10 · 2026-09-27
- GPT-6 Astra reportedly proves weak Goldbach case unconditionally in 2 pages — yshan2u · 2026-09-27
- Dean Ball: AI safety and e/acc are natural allies — drop the kayfabe — deanwball · 2026-09-27
- Auto-research agents are coming, and peer review may shatter without defenses — askerlee · 2026-09-27