Jeremy Howard

Jeremy Howard looked at the state of deep learning education in the mid-2010s and saw a problem that offended him: the most transformative technology of our era was locked behind a gate of PhD prerequisites, expensive compute, and impenetrable academic jargon. His response was fast.ai -- a free course and software library designed to prove that you do not need a doctorate in mathematics to build world-class AI systems. He was right, and the impact has been extraordinary.

The fast.ai course, "Practical Deep Learning for Coders," takes a top-down approach that was radical when it launched. Instead of starting with calculus and linear algebra, Howard starts with working code. Students build image classifiers, text models, and recommendation systems in their first lessons, and only then begin peeling back the layers to understand the mathematics and theory underneath. It is an approach borrowed from how people actually learn most skills -- by doing first, then understanding -- and it has proven devastatingly effective. Hundreds of thousands of students worldwide have taken the course, and many have gone on to publish research, build companies, and contribute to the field.

The fastai software library embodies the same philosophy in code. Built on top of PyTorch, it provides high-level abstractions that let developers train state-of-the-art models in a fraction of the code that would otherwise be required. But unlike some high-level libraries that sacrifice flexibility for simplicity, fastai is designed so that every layer of abstraction can be peeled back when needed. It is opinionated but not rigid, making it equally useful for rapid prototyping and serious research.

Howard's impact extends beyond his own work. He has been a vocal advocate for AI safety, responsible deployment, and the importance of diverse perspectives in AI development. His argument is that democratizing AI is itself a safety measure -- that concentrating the technology in the hands of a few large corporations is more dangerous than spreading the knowledge broadly. Whether or not one fully agrees, the premise has inspired a movement of practitioners who believe that open, accessible AI education is a public good worth fighting for.