Can AI make mathematics more human?

Mathematician Yang-Hui He discusses how artificial intelligence is revolutionizing mathematical work and argues that this is a positive development. He explains that machine learning, driven by advancements in deep learning architectures, has surged in recent years, catching mathematicians off guard. He recalls how he, initially a string theorist, pivoted to understanding machine learning after witnessing its rapid growth in 2017.

Yang-Hui applied a simple neural network to Calabi-Yau manifolds, a central concept in string theory, and was astonished to find the network could predict topological properties with remarkable accuracy.

He notes that machine learning could greatly aid string theory, particularly in identifying the correct version that describes our world by efficiently searching through countless Calabi-Yau manifold possibilities. However, Yang-Hui later moved away from string theory, realizing that he was using machines to explore the structure of mathematics in a broader sense.

He emphasizes that physicists are more open to using AI due to their familiarity with computational tools and large datasets, while mathematicians remain skeptical. Despite this, Yang-Hui believes AI will eventually change the way mathematicians communicate, acting as a "traveling salesman" to persuade various mathematical fields to embrace data-driven methods.

He predicts that AI will make mathematics more human by fostering a common language among diverse mathematical disciplines. Yang-Hui has already utilized AI to uncover patterns in data and formulate new hypotheses, but his ultimate goal is to leverage the technology to solve significant open problems. He believes AI's rapid improvement, as demonstrated by projects like FrontierMath, will bring this goal closer to reality.