Training Sand to Think: Artificial General Intelligence & Future of Physics
Research Lead, Blueshift Team, Google DeepMind; Former Physics Faculty, Princeton and Stanford
Adam Brown reviews the rapid advancements in large language models, discussing their enhanced scientific reasoning capabilities and speculating on the implications for future research trajectories within theoretical physics.
Our civilization has learned how to turn sand into silicon chips, silicon chips into neural networks, and neural networks into Artificial Intelligences (AIs). Over the last half-decade, the capabilities of large language model AIs (like ChatGPT and Gemini) have leapt from babbling preschoolers to International Math Olympiad gold medalists, and now beyond. This talk reviews recent progress in training AIs to do science and reasoning, and speculates as to what it will mean for the future of physics if these trends continue.
About the Speaker