MIT FutureTech's Science Task Taxonomy: 12 top-level areas, 2,433 sub-areas, 208,202 tasks
JMateosGarcia · x · 2026-10-06
MIT FutureTech researchers (Emmens, Liu, Trišović, Neil Thompson) released Scientific Work, introducing the Science Task Taxonomy: a framework mapping scientific work across 232 subfields in three levels — 12 Level-1 task areas, 114 Level-2, and 2,433 Level-3 — down to 208,202 representative tasks. Built via creative method recombination (feature classification, clustering, LLM labelling; network visualisation by Gemini). Findings: scientific work is more cognitive, involves fewer interpersonal interactions, is less manual, less codifiable, and the median scientific task takes over two hours longer than economy-wide tasks. The task-level measurement has implications for understanding AI's effect on science. Paper and taxonomy data are publicly available.
Related event: MIT Releases Science Task Taxonomy Covering 208,000 Tasks(2 posts)→
More from AGI Musings
- AI Ethicist Carissa Veliz's 'Prophecy' Shortlisted for FT Best Business Book Award — CarissaVeliz · 2026-10-06
- Rumor: OpenAI to publish ~400 AI-generated proofs with major math breakthroughs — Dr_Singularity · 2026-10-06
- AI hyperscalers near $500B in debt this year as borrowing spree pushes rates to 2002 highs — churchkey · 2026-10-06
- Nobel laureate Shiller: AI doom narratives could become a self-fulfilling economic prophecy — IgorCarron · 2026-10-06
- Researcher: AI welfare field standards are a joke, funding flows to pro-consciousness groups — erikphoel · 2026-10-06
- Why AI agents stay useless at home: models can't tell what actually matters — jdjohnson · 2026-10-06