AI is rapidly making it possible to automate increasingly large parts of scientific work. Much of that progress is genuinely exciting. But it also raises a deeper question.
Science is more than a sequence of tasks. It is a social system built on diversity of ideas, accountability, criticism, replication, and the willingness to be proven wrong.
In a recent Science editorial, Brian Uzzi and I argued that fully autonomous “end-to-end science” could fundamentally alter the evolutionary dynamics that make science both reliable and creative.
Our new American Scientist article develops that argument further. We ask what happens when we mistake the machinery of scientific production for science itself — and why efficiency alone cannot substitute for the institutions and norms that have made science one of humanity’s greatest engines of discovery.
The question is not whether AI will transform science. The question is whether, in doing so, it preserves the conditions for scientific discovery rather than merely accelerating scientific production.
