The Periodic Table of Machine Learning: Discovering Structure in the Chaos of AI

About this session

What if machine learning had its own periodic table, not of atoms but of algorithms? In this talk, I’ll share how our team of researchers from MIT, Microsoft, and Google discovered a single equation that unifies over 20 foundational AI methods, from clustering to contrastive learning, into one simple, elegant framework. Just as the periodic table revealed hidden patterns in chemistry, our "Periodic Table of Machine Learning" uncovers deep connections between seemingly unrelated algorithms and even reveals entirely new ones waiting to be explored. Beyond theory, this perspective enabled us to build unsupervised AI systems that classify images with no human labels, outperforming previous state-of-the-art methods. It’s a story of simplicity beneath complexity, and how uncovering structure can spark breakthroughs.

Speakers

Shaden Alshammari

AI Research, MIT