Artificial intelligence is fundamentally changing the pace and scale of scientific discovery, leading to breakthroughs in fields as varied as biology, genomics, materials science and mathematics, Google Deepmind executive Pushmeet Kohli told a standing-room only crowd in the Friend Center on May 2.

This ambitious approach to solving the world’s most pressing issues is at the heart of DeepMind’s mission to build AI to benefit humanity, said Kohli, who is vice president for science and strategic initiatives for the company.

“We believe that one of the ways we can take our civilization forward is to really understand about nature, and the best way we can do that is through science,” he said.

Kohli was the fifth and final Distinguished Lecture organized by the Princeton Laboratory for Artificial Intelligence this semester. The series, which will continue in the fall, brought scholars to campus whose research demonstrated the transformative impact AI could have across disciplines. 

Kohli is “one of the premier experts in artificial intelligence today, especially artificial intelligence for science,” according to Sanjeev Arora, director of Princeton Language and Intelligence, which hosted the event. After earning his doctorate in computer vision from Oxford Brooks University, Kohli spent a decade at Microsoft Research, where he was most recently director of research for its cognition group.

He joined DeepMind in 2017, soon after the historic five-game Go match where a top player was defeated by AlphaGo, a program developed by the company. The company went on to release AlphaFold, a groundbreaking AI system that predicts the 3D structure of proteins and won the 2024 Nobel Prize in chemistry.

DeepMind doesn’t organize itself by topics, Kohli said, but by specific goals that, if solved, would have “transformative impact” on a given field. There must also be scientific consensus that solving the problem is possible, he said, but it would take at least ten years. DeepMind’s researchers will take it on if they believe they can solve it in half or a third of that time, he said.

“Our approach is very different,” he said. “We are focused on achieving those goals with a team that is solely looking at that goal.”

In his talk, Kohli discussed DeepMind’s key breakthroughs in different scientific disciplines, including AlphaFold’s impact on structural biology, as well the AI methodologies behind these projects, and the company’s vision for the future of AI.