Toward autonomous science with agentic artificial intelligence.

Agentic AI systems like Co-Scientist, Robin, and Biomni can generate hypotheses, design experiments, analyze data, and refine reasoning through feedback, advancing autonomous scientific reasoning.

• Why it matters: Understanding how AI can autonomously contribute to scientific discovery addresses a critical gap in enabling scalable, efficient research processes while managing risks like hallucination and bias.
• What they did: The authors review recent AI systems in biomedicine, discuss scaling laws and autonomous labs, and explore challenges such as irreproducibility and dual use, emphasizing responsible collaboration.
• The result: This work highlights AI’s potential to transform science through autonomous reasoning, contingent on addressing key challenges and fostering responsible human-AI partnerships.