On September 3, a collaborative team of researchers from Google, HHMI Janelia, and other institutions unveiled MaleCNS, a comprehensive mapping of the male fruit fly's connectome. This detailed map illustrates the entire central nervous system, encompassing the brain and ventral nerve cord, featuring a staggering 166,700 neurons and 125 million synapses. Just a day later, Georgia Tech graduate student Evan Sinclair Smith connected MaleCNS to Minecraft. In this setup, the game's events are fed into the sensory neurons of the fly, while the outputs from the network's neurons are utilized to control the character. The virtual fly navigates its environment by searching for food, responding to light, and avoiding other players. This innovative demonstration quickly gained popularity, leading enthusiasts to have the digital fruit fly engage in various activities, including playing Doom, Beat Saber, driving vehicles, and even participating in cryptocurrency trading. However, there is an important caveat: these demonstrations heavily rely on the developers' choices regarding which game events correspond to specific sensory neurons and how output neurons relate to control mechanisms. The movements are pre-programmed, with the neural network simply determining which action is activated. Meanwhile, discussions emerged on Reddit about the ethical implications of these experiments. Some users even created a "heaven" for the fly—a virtual space where it could freely explore without any obligations. It is crucial to note that these experiments are far from achieving a true "digital clone" of a fly. Although MaleCNS provides valuable insights into the wiring and neurotransmitter information of the nervous system, it does not account for the physiological characteristics of neurons and synapses, receptor dynamics, neuromodulation, or other essential processes necessary for a complete biological simulation. Neuroscientist Patrick Mineault cautions against overinterpreting these captivating demonstrations, stating, "Virtual fly brains are not conscious." Whether the virtual fly is maneuvering in Minecraft or controlling a drone, the underlying neural network essentially replicates the layout of a fly's nervous system. However, substituting the "fly wiring" with a random neural network would yield a similar outcome. So, what is the significance of this work? Connectomes can provide scientists with insights into the relationship between neural circuits and behavior. By manipulating specific cells in a model, researchers can make predictions and then verify those predictions in live flies. For instance, in 2024, a study utilizing a model based on a previous connectome artificially stimulated sugar-sensing neurons, resulting in activity that spread through the network until it reached a motor neuron responsible for the proboscis. This predicted outcome was later confirmed in actual flies. However, a complete simulation of a fly remains a distant goal, as even in that study, only a few hundred neurons were activated out of approximately 140,000. Mineault suggests that to truly determine if a "digitized" fly could be considered a replica of its biological version, researchers should seek behaviors that are specific to individual flies. For example, could a virtual fly independently navigate a maze that a real fly had learned during its life?
Informational material. 18+.
