From Machine Learning to the Nature of Humans: A Physics Perspective

By now, most of us have used ChatGPT to write something, been amazed by images generated with DALL·E, or received the exact recommendation we were looking for while shopping online. These tools seem incredibly powerful, but can we use them to help us do science? The answer is yes, and researchers are already doing it: from forecasting the weather to predicting how proteins fold, and even running chemistry labs. The ultimate goal? Creating machines that can make scientific discoveries on their own. In this talk, I’ll explore how close we really are to that goal and what it will take to get there.

What makes something a particle? ‘Elementary' particles like the electron can not be divided into smaller units. But many elementary particles were only discovered by studying the behavior of larger 'composite' particles, such as protons, in the 'particle zoo'. In fact, from planets to pollen grains, physics has a long and productive history of learning about things by studying them as if they were truly indivisible particles. What happens when we apply this approach to organisms? When we attempt to analyse the movements of humans and animals as if they were particles, we learn something about the physical nature of agency and the small, invisible forces that make us into individuals.