Neuron–astrocyte networks might perform the core computations performed by AI transformer blocks

Fascinating ideas from Kozachkov et al. Their text contains primers on Astrocyte biology and the transformers found in AI Generative Pre-trained Transformers such as ChatGPT.  

Significance

Transformers have become the default choice of neural architecture for many machine learning applications. Their success across multiple domains such as language, vision, and speech raises the question: How can one build Transformers using biological computational units? At the same time, in the glial community, there is gradually accumulating evidence that astrocytes, formerly believed to be passive house-keeping cells in the brain, in fact play an important role in the brain’s information processing and computation. In this work we hypothesize that neuron–astrocyte networks can naturally implement the core computation performed by the Transformer block in AI. The omnipresence of astrocytes in almost any brain area may explain the success of Transformers across a diverse set of information domains and computational tasks.

Abstract

Glial cells account for between 50% and 90% of all human brain cells, and serve a variety of important developmental, structural, and metabolic functions. Recent experimental efforts suggest that astrocytes, a type of glial cell, are also directly involved in core cognitive processes such as learning and memory. While it is well established that astrocytes and neurons are connected to one another in feedback loops across many timescales and spatial scales, there is a gap in understanding the computational role of neuron–astrocyte interactions. To help bridge this gap, we draw on recent advances in AI and astrocyte imaging technology. In particular, we show that neuron–astrocyte networks can naturally perform the core computation of a Transformer, a particularly successful type of AI architecture. In doing so, we provide a concrete, normative, and experimentally testable account of neuron–astrocyte communication. Because Transformers are so successful across a wide variety of task domains, such as language, vision, and audition, our analysis may help explain the ubiquity, flexibility, and power of the brain’s neuron–astrocyte networks.