What the fly sees
Live balance of the direction-selective T4/T5 cells that prefer leftward vs rightward motion.
Loading the fly's brain…
The reply is a template filled in with numbers measured from the simulation. The fly can't read. Its eye is far too coarse to make out letters, and the model has no language areas. Only the facts are its own.
Inside the brain
Every square is one neuron, drawn where its column sits in the eye. Each panel is one cell type (all cells of a type share the same wiring). Click a panel to see its wiring and activity.
What is this?
The model
This is flyvis, the model from Lappalainen et al., Connectome-constrained networks predict neural activity across the fly visual system (Nature, 2024). Its wiring comes from electron-microscopy reconstructions of the fruit fly's optic lobe. Training only tuned one strength per connection type, plus each cell type's resting potential and time constant, so that the network could report visual motion. The network running here is model flow/0000/000 from the published ensemble.
The simulation
Each neuron is a leaky, non-spiking unit and the synapses are graded, as in flyvis. The email is drawn onto a strip that scrolls across the eye, sampled by the same box filter flyvis uses, and integrated in steps of 20 ms. This JavaScript port is tested against flyvis itself and matches it to within about 10-6.
What it can't do
This is only the visual system: the part that turns light into motion signals. There is no memory, no language and no decision-making here. The fly's reply tells you how its neurons responded, not what the email said.
Model, weights and connectome: flyvis, MIT License © 2023 Janne K. Lappalainen, Fabian D. Tschopp, Mason McGill, Jakob H. Macke, Srinivas C. Turaga. This site isn't affiliated with the authors.