Research News & Updates

OpenAI's new model has produced an original proof disproving a prominent open problem in geometry, as claimed by the company. This achievement marks the first time AI has autonomously solved a central problem in mathematics. The model's capabilities have broad implications for various fields, including biology, physics, and medicine.

Customizing AI agents requires various techniques, including prompt engineering, retrieval-augmented generation, agent tool and skill injection, supervised fine-tuning, parameter-efficient fine-tuning, direct preference optimization, and reinforcement learning.

MIT Associate Professor Connor Coley uses AI to help identify compounds that could make good drug candidates, developing models like ShEPhERD and FlowER.

A new study by MIT labor economist David Autor finds that new job creation is often driven by urban settings, young and educated workers, and new forms of expertise linked to technology.