John Hopfield: Physics View of the Mind and Neurobiology

Lex introduces John Hopfield, a Princeton professor whose life's work wove through biology, chemistry, neuroscience, and physics—seeing the messy world of biology through a physicist's piercing eyes. Best known for associative neural networks now called Hopfield networks, an early catalyst of modern deep learning, he won the 2019 Franklin Medal for applying theoretical physics to biological questions with impact on machine learning. As his 2018 article "Now What?" notes, his accomplishments came by asking that question and often changing direction.

After the podcast's usual appeals and a Cash App sponsor read praising its fractional-share order-execution algorithm, Lex turns to the conversation.

Asked what difference between biological and artificial neural networks most captivates him, Hopfield points to how neurons have many components and properties, and how evolution sharpens any useful quirk in a molecule or cell into a feature rather than a glitch—an aspect completely suppressed in artificial neural networks. He confirms glitches can indeed become features in biology.

Hopfield illustrates with oscillators that phase-lock, citing the Millennium Bridge over the Thames (2001): pedestrians walking near the same frequency fell into step with the bridge's slight sway until it oscillated, because engineers missed the side-to-side motion of walking. Likewise loosely coupled nerve cells firing action potentials can lock together under some circumstances—a computational feature evolution can exploit, whereas most artificial networks lack action potentials entirely.

Asked how evolution leverages biology's weird mess, Hopfield describes DNA duplication: one copy keeps coding the old protein while the other drifts to code something slightly different and useful, letting evolution slowly improve a new function while retaining the old. He draws a playful dual with companies—whose "evolution" means some close and others open—and notes IBM built its first PC only by creating an isolated unit and changing its culture, while biology finds new markets far more adaptively.

Asked the most beautiful aspect of the mind, Hopfield says adaptation is everything, distinguishing learning over evolutionary generations from learning within one individual's lifetime—biology has both. What makes neurobiology hard, he says, is that it is a mathematical system built atop this other evolutionary system.