‘AI Scientists’ Are Improving, but What Are the Fundamental Limits?

‘AI scientists’ are improving, but what are the fundamental limits?

AI scientists have improved leaps and bounds in the last few years but there are still some significant concerns and limitations.

For AI co-scientists to scale, scientists have to trust them. The architectural bets to earn it vary.

A growing number of vendors, from Google DeepMind to Benchling to traditional ELN companies like Sapio Sciences, are pitching “co-scientist” functionality to bench researchers. They are part of a multibillion-dollar wave of AI life-sciences deals that has accelerated since late 2025. To cite but a few examples, OpenAI introduced GPT-Rosalind and named Amgen, Moderna, and… The post For AI co-scientists to scale, scientists have to trust them. The…