Cristian Ponce

What happens when AI systems stop being tools and begin acting like collaborators in scientific thought? Multi-agent architectures such as SciAgents and Google’s “AI Co-Scientist” are pioneering hypothesis generation by dividing scientific reasoning into specialized sub-agents: literature retrievers, causal mappers, and graph-based reasoners. Unlike single models, these teams of agents mimic the structure of scientific collaboration itself — brainstorming, critiquing, and refining ideas. In synthetic biology, such systems could propose new gene circuits, uncover hidden regulatory logic, or suggest underexplored protein folds. This session asks: how far should we trust AI-generated hypotheses, and how do we validate them responsibly? With machine-driven insight now on the horizon, the very architecture of discovery may shift — from lone researchers and teams of humans to networks of humans and machines co-creating the future of biology.