In this episode of the Google DeepMind podcast, Pushmeet Kohli, the head of scientific research, explores how AI is changing the landscape of various scientific disciplines. He highlights groundbreaking innovations like AlphaFold3, which predicts the structures of biomolecules beyond just proteins, thus transforming the field of biology. He also discusses GraphCast, a new model that enhances the accuracy and efficiency of weather forecasts, and GNOME, an AI tool that forecasts millions of stable, undiscovered inorganic materials. This could lead to significant advancements in battery technology and superconductivity. Kohli points out the importance of tackling fundamental research problems with accessible data and effective evaluation metrics. He demonstrates how AI can expedite scientific discovery by managing complex datasets and uncovering hidden patterns. The conversation also underscores AI's dual role in analyzing existing data and generating fresh hypotheses, illustrated by the FunSearch project, which achieved a novel result in computer science.
Sign in to continue reading, translating and more.
Continue