Astrocytes: neuroscience and AI implications

[ionelv]

Quanta published yesterday a very good summary of the state of our understanding how the brain signaling and modulation works, mentioning 1989 foundational astrocyte research, a pivotal research paper published in 2016 about astrocytes, a few revelatory 2025 astrocyte papers and the summary is damning:

“Neuroscience has only cared about neurons for a century now, and we don’t yet have a cure for a single brain disorder,” Papouin said. The way to change that, he said, is to accept the existence and influence of non-neuronal cells such as astrocytes, and to include them in models and experiments.

Most neuroscientists haven’t received that memo, Freeman said. “Ninety-nine percent of people who are out there doing experiments on circuits don’t even think about what the astrocyte might be doing. And it could have really profound effects on how that circuit functions.”

More interestingly is how our current AI revolution is including astrocytes into its models (see ANGNs): none of the publicly available models make use of these more accurate models.