I connected a fruit fly connectome to tic-tac-toe (with a minimax safety net)

Description and use

The project shows how real fruit fly connectome data can be mapped to a tic‑tac‑toe agent, using a minimax safety net to improve play.

Problem

How to use a real fruit fly neural circuit to play tic‑tac‑toe.

Outcome

The project demonstrates that a real fruit fly connectome can be mapped to a simple game, producing a playable tic‑tac‑toe agent. The fly’s circuit alone tends to lose, but when combined with a minimax safety net it can play competitively.

Workflow

• Select neurons that send motor commands and their central brain inputs.
• Use connectome data to obtain neuron‑to‑neuron connections, synapse counts, and neurotransmitter predictions.
• Map each of the 9 tic‑tac‑toe squares to a group of output neurons and each board state to a group of input neurons.
• Simulate signal flow through the network using the real wiring and synapse strengths.
• Determine the fly’s preferred move as the output group with the strongest activity.
• Run minimax to generate optimal moves; let the fly pick among minimax‑recommended moves.
• If minimax is disabled, run the raw circuit alone to observe its behavior.
• Play the game via terminal interface.

Evidence

Scientists have mapped the whole nervous system of a fruit fly: every neuron and every connection, in a public database called a connectome.

Source

Python, neuprint-python, networkx and pandas do the work.

Source