REDDIT · REDDIT// 8d agoRESEARCH PAPER

MIT study douses AI job apocalypse

MIT FutureTech’s latest paper evaluates more than 3,000 labor-market tasks with 17,000 worker judgments and finds AI automation looks more like a gradual “rising tide” than a sudden wave of layoffs. The study says text-heavy tasks are getting increasingly doable for models, but reliable near-perfect output is still years away.

// ANALYSIS

This is a useful antidote to apocalypse framing: AI is spreading fast, but the binding constraint is still quality, integration, and workflow redesign, not instant mass replacement.

–The methodology is stronger than benchmark-chasing because workers judged outputs against real job usefulness, not just synthetic test scores
–The numbers point to steady encroachment: roughly 50% task success in 2024-Q2, about 65% by 2025-Q3, and a plausible 80% to 95% by 2029 at minimally sufficient quality
–The biggest exposure is in text-heavy, information-processing work, while legal and precision-sensitive roles still resist full automation
–“Good enough” AI can still reshape headcount over time, even if it does not trigger a sudden job cliff
–The study is limited to text-addressable tasks, so it likely undercounts physical, multimodal, and on-the-ground labor dynamics

// TAGS

researchllmautomationcrashing-waves-vs-rising-tides-preliminary-findings-on-ai-automation-from-thousands-of-worker-evaluations-of-labor-market-tasks

DISCOVERED

8d ago

2026-04-03

PUBLISHED

8d ago

2026-04-03

RELEVANCE

8/ 10

AUTHOR

Anen-o-me