MIT's Sasha Rakhlin Urges Universities to Invest Now as AI Nears Human-Level Research Work
Updated
Updated · MIT News · Oct 8
MIT's Sasha Rakhlin Urges Universities to Invest Now as AI Nears Human-Level Research Work
2 articles · Updated · MIT News · Oct 8
Summary
Rakhlin argues universities should prepare for a near future in which AI outperforms humans in many intellectual tasks, especially across mathematics, statistics, machine learning and engineering.
A key driver is fast verification: formal proofs can be checked automatically and code can be tested quickly, letting models iterate, learn from outcomes and accelerate their own improvement.
That shift weakens polished papers as proof of individual expertise, he says, pushing departments to reward question-setting, replication, negative results, shared datasets and clear responsibility for AI-assisted work.
Graduate training also needs redesign because offloading routine calculations, coding and failed attempts to AI can strip away the formative work through which students build intuition and judgment.
MIT and peers should invest now in shared AI infrastructure—compute, secure data systems and lab workflows that capture hypotheses, failures and interpretations—to connect laboratories while preserving consent, credit and technological independence.