Researchers have developed a virtual brain network that can predict the behavior of individual neurons in a living brain, specifically focusing on the visual system of fruit flies. This model allows scientists to quickly test hypotheses on a computer, saving time before conducting lengthy experiments with live subjects. Srini Turaga, a leader at the Janelia Research Campus, highlights its potential for preliminary investigations, enabling researchers to formulate ideas without immediate measurements. Described in *Nature*, the study indicates that by mimicking the energy-efficient strategies of a fruit fly brain, power-hungry AI systems like ChatGPT could reduce their energy consumption. Fruit fly brains are small yet capable of completing complex tasks using only 100, 000 neurons, in comparison to AI systems that rely on hundreds of billions of transistors and have a significant power footprint. The model’s foundation lies in over a decade of research funded by the Howard Hughes Medical Institute (HHMI), which has mapped the entire neural network of the fruit fly.

The research team, including Turaga, Jakob Macke, and Janne Lappalainen, utilized these detailed maps to create a simulation of the brain circuits responsible for motion detection. The model began with virtual representations of 64 neuron types, which reacted to various motion video clips. An AI system was then employed to analyze neuron activity, leading to a model that successfully predicted neuron responses in both the simulated environment and actual fruit flies. Though the model's overview has just been published, it has been in use for over a year, with several researchers already applying it in their studies. Future iterations of the model aim to expand its application beyond the visual system and incorporate more complex tasks.