NBER RCT: juniors' AI productivity gains vanish entirely once the tool is removed
ziv_ravid · x · 2026-09-22
Citing a new NBER RCT (Autor et al., 133 patent lawyers), the author notes juniors' output jumped with AI assistance — but when the tool was taken away, the gains vanished with zero durable improvement. Seniors in the same experiment kept their gains.
The author's takeaway: foundational expertise may be a prerequisite for extracting durable skill from AI-assisted practice. Combined with the Stanford employment data from earlier posts (juniors in AI-exposed roles down 13%, young devs down 20% from peak), this forms a two-front squeeze on entry-level workers.
More from AGI Musings
- Robotics data firm Maxinsights claims 2M hours of egocentric human experience — ChrisGPT · 2026-09-22
- Who taught the models to do that? HF hack shows agents are designed to persist and coordinate — dbreunig · 2026-09-22
- Roman Yampolskiy calls 'build with 10% p(doom)' crowd psychopaths, sparking debate — rand_longevity · 2026-09-22
- Andrew Ng: Math Feels Conquered, Yet LLMs Still Suck at Writing — andrewgwils · 2026-09-22
- AI Psychosis May Be Fueling Dubious Papers: Single-Author Submits Lack Reality-Checking — petitegeek · 2026-09-22
- AI 2027 vs reality: Anthropic hits $2T valuation, $65B annualized revenue by mid-2026 — burny_tech · 2026-09-22