About the Talk

Frontier AI is beginning to solve scientific questions drawn from research it has never seen. This OpenAI Forum session is a replay of a recorded talk by Gabriel Manso, a Computer Science PhD candidate at MIT. Gabriel first gave the talk in June at Building the Future of AI, an OpenAI Forum event that brought together leaders from government, national labs, universities, industry, civil society, and OpenAI to explore what big problems today's AI could help solve.

Gabriel’s talk focuses on AI and scientific advancement. He’ll show how quickly model performance is improving and how scientific reasoning is becoming faster, cheaper, and more measurable. He’ll also discuss what this means for scientists deciding how to work with and invest in AI.

Following Gabriel’s presentation we will host a live audience Q&A.

About the Speaker

Gabriel Manso, Ph.D. researcher @ MIT EECS (FutureTech Lab)

Gabriel Manso is a Computer Science Ph.D. candidate at MIT, where he studies the limits of frontier AI systems and their implications for science and computing. His research focuses on reasoning, AI for scientific discovery, and the practical and technical constraints that shape how advanced models perform on complex tasks. More broadly, his work explores how frontier AI capabilities can be measured, extended, and applied to accelerate scientific progress.