Nina Miolane | The Shape of Intelligence

Why It Matters

The shared toroidal manifold hints at a universal principle linking brains and AI, guiding more biologically inspired algorithms and accelerating neuroscience breakthroughs.

Key Takeaways

• Neural activity maps onto a 150‑dimensional torus shape.
• Both biological and artificial networks converge to similar toroidal geometry.
• Torus pattern appears across species: mice, rats, monkeys, humans.
• AI training replicates evolutionary neural structures in minutes.
• Suggests universal laws governing brain and machine intelligence.

Summary

The video presents recent findings that the collective firing patterns of 150 recorded neurons can be represented as a point in a 150‑dimensional space, revealing a toroidal geometry that appears to be a fundamental shape of neural computation.

Researchers showed that both biological brains and deep‑learning models, despite vastly different training times—millions of years of evolution versus minutes of gradient descent—converge on the same torus structure. The pattern persists across network initializations, architectural variations, and across species from rodents to primates.

As the speaker notes, “you crack open the artificial neural network and you print the activity… you do get that torus.” The torus has been documented in mice, rats, to a lesser extent in monkeys and humans, suggesting a reproducible signature of high‑dimensional neural dynamics.

If neural computation obeys universal geometric constraints, AI designers could exploit toroidal representations to build more efficient, robust models, while neuroscientists gain a quantitative bridge linking brain activity to machine learning theory.

Original Description

Researchers mapped the collective activity of neurons in humans and AIs, and the shape converged on the torus. Could there be a universal geometry of intelligence? Nina Miolane, Assistant Professor at UC Santa Barbara and Director of the Geometric Intelligence Lab, explains in her Long Now Talk with science historian Claire Isabel Webb.

Watch the full episode in the link below. Find more videos on long-term thinking and planetary philosophy on the Long Now channel.