We are on the verge of a scientific revolution in AI. But no one really knows, "Why AI Works". In this talk, we discuss the results of a 10-year research program to peer inside the details of AI models, and understand the emerging signatures of complexity in these models that look exactly like the same signatures seen in experimental measurements on real-world neurons...like our own brains This talk is based on my research, done in joint with UC Berkeley, developing a new theory of AI. My research has been published in top science journals like Nature Communications and JMLR (Journal of Machine Learning Research) and has been featured at top AI conferences including ICML, KDD, and NeurIPS. I am also a regular invited speaker at top universities like UC Berkeley and Stanford.

I have personally pioneered this research, funding it through my consulting practice. I have also developed a supporting open-source tool, weightwatcher, that has over 100 downloads and is widely used around the world to help build modern AI models. This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at https://www.ted.com/tedx