Scientists work with virtual AI teams

Here's a fun piece: I had a chat with some scientists who have used teams of virtual AI 'agents' to help them formulate new experimental ideas, critique their papers, and run lab meetings to think through different aspects of their work. 

All felt that the adoption of LLMs into idea generation and experimental design is as inevitable as the adoption of Internet searches into science. But they differed in whether they felt the results would be dramatic or always beneficial. Chatbots might make things more efficient and help guide graduate students, for example, but they might also undercut learning, or take away the "fun part" of science, such as coming up with new ideas.  

My story for Nature: https://www.nature.com/articles/d41586-025-02028-5