In Defense Of Dynamics
NeuroAI is all the rage right now, and for good reason. LLMs, ChatGPT, and now the plethora of AI chatbots really are a major breakthrough. Still, there is clearly a lot we can still learn from the brain. AI is trained on the entire internet and uses huge amounts of power. Yet the human brain learns quickly and runs on the power of a single donut. So obviously, there is much more to learn from the brain with respect to AI.
That being said, I really feel that there is more to learn about the brain that just isn’t about translating it for AI. I want to outline one of those reasons here. Non-ML tools such as dynamics are useful for understanding the brain, and are still worth learning, improving, and using, even if they don’t directly or indirectly further the neuroAI mission.
Dynamics is, at its core, the mathematics of change. Neurons fire, voltages fluctuate, neurotransmitters orchestrate activity across ensembles of cells. And in the end, a computation is done. We neurodynamists use the mathematics of change to understand how this orchestra works. We aim to understand what pieces of the brain give rise to what functions. This is broader than the classical debate between reduction and holism that gave rise to systems biology. Dynamics takes the reduced pieces neuroscientists give us and uses them to generate the emergent behavior that holism is built around. In the end, we are left with a mechanistic description, one we can actually understand. There is no black box. Change the pieces, and the behavior changes.
This is more than intellectual curiosity. It gives us a framework for what happens when the system breaks. The brain is a remarkably robust machine, but when it fails, specific components (genes, proteins, ion channels) are altered or missing. These small subtle changes can cascade into full blown disease. Dynamics gives us a language to translate from a broken protein to a disease.
Take, for example, seizures. An underactive potassium channel can lead to an overactive, overexcitable brain. Dynamics gives us the language to understand the full cascade; how a potassium channel dysfunction can lead to seizures. But this is more than idle curiosity. This is a key reason pharma companies invested in Retigabine, a potassium channel opener that helps correct this imbalance. Dynamical and computational models converged on the idea that upregulating slow potassium currents could put the brakes on seizures. These models didn’t exist in isolation. Nonetheless, they explained experimental observations and pointed toward a clear, investible, intervention. This confluence of neuroscience and dynamical thinking helped drive the development of retigabine. This demonstrates that dynamics is not just useful for understanding how the brain breaks, but for figuring out how to fix it.
This is far from an isolated case. From tremor to depression, dynamics provides a framework for describing how disease unfolds, what is broken, and how those failures produce symptoms from the instruments in the orchestra of the brain. So as we rush toward neuroAI, don’t forget the importance of dynamics in unraveling the mysteries of the brain. It is a complementary tool; one that can uncover mechanisms and, ultimately, help treat disease.
Author: Alexander J. White
Reference:
Knowles, W. D., Traub, R. D., Wong, R. K., & Miles, R. (1985). Properties of neural networks: experimentation and modeling of the epileptic hippocampal slice. Trends in Neurosciences, 8, 73-79.
Pan, M. K., Chen, L. Y., Wang, Y. M., White, A., Ho, J. Y., Chen, S. Y., ... & Ermentrout, G. (2026). Circuitry dynamics of the cerebellum inform differential therapeutic responses and patient stratification in essential tremor.
Charlton, C. E., Karvelis, P., McIntyre, R. S., & Diaconescu, A. O. (2023). Suicide prevention and ketamine: insights from computational modeling. Frontiers in psychiatry, 14, 1214018.
留言
張貼留言