Terrence Sejnowski Chats Chatbots and What We Can Learn at the Intersection of AI and Neuroscience

Why It Matters

Sejnowski’s insights illustrate how brain‑inspired AI can accelerate scientific discovery while highlighting ethical challenges that require interdisciplinary oversight, shaping the future trajectory of both fields.

Key Takeaways

• Sejnowski pioneered the Boltzmann machine, enabling multi‑layer learning.
• Early speech‑to‑text work foreshadowed modern chat‑bot capabilities in industry.
• Computational neuroscience bridges brain biology and deep‑learning architectures.
• Neuro‑AI leverages brain‑like models to improve AI interpretability.
• Philosophical input is essential to address AI bias and hallucination.

Summary

In this Beyond Lab Walls episode, computational neuroscientist Terrence Sejnowski discusses the convergence of artificial intelligence and brain science. He recounts his journey from a childhood volcano experiment to pioneering the Boltzmann machine—a learning algorithm that made multi‑layer neural networks biologically plausible—and early speech‑to‑text systems that prefigured today’s chat‑bots. Sejnowski explains how computational neuroscience evolved from simulating tiny neuron populations to the modern field of neuro‑AI, where deep‑learning architectures mirror cortical circuits. He highlights the 2012 breakthrough of convolutional neural networks, which not only outperformed hand‑crafted vision systems but also offered a window into monkey visual cortex responses. The conversation also touches on scaling challenges: early networks ran on computers a billion times slower than today’s phones, yet the exponential growth in compute power has unlocked models capable of generalizing across unseen data. Memorable moments include Sejnowski’s analogy that the brain “regulates its ability to hallucinate,” a problem mirrored in large language models that fabricate plausible‑sounding facts. He stresses the need for philosophers and ethicists to guide AI development, noting that bias and hallucination stem from the same data‑driven processes that shape human cognition. The interview underscores a dual agenda: using brain‑inspired AI to solve complex tasks faster while simultaneously leveraging AI to decode neural function. This interdisciplinary loop promises more transparent models, better tools for neuroscience, and a framework for responsibly governing emerging AI systems.

Original Description

Terrence Sejnowski, PhD, is a professor, head of the Computational Neurobiology Laboratory, and holder of the Francis Crick Chair at Salk, and he's played a hand in two entirely new scientific fields emerging—first computational neuroscience, then NeuroAI. Hear what it was like creating the first text-to-speech program, best practices for using modern chatbots, and how computer models are helping neuroscientists understand the human brain.