An experiment that combines nervous system simulation with encrypted transactions has recently drawn attention. Coinbase Software engineer Alex Wormuth integrated a digital simulation system based on the neural connection structure of fruit flies into the Bitcoin market and allocated $100 to it for trading. One day after the experiment was launched, the system had a profit of $1 on paper.

Connect to the simulation system of Coinbase.

This system is named Stonkfly. It does not involve real fruit flies in transactions; instead, it connects a simulated model of the nervous system of adult male fruit flies to the public market data of BTC-USDC within Coinbase.

The base model used in Wormuth comes from the recently released MaleCNS, v1.0, and connectome. This model attempts to replicate the neural connections within the fruit fly's brain and central nervous system, and then maps these connections to external input and feedback mechanisms.

Trading signals come from vision and feedback.

Stonkfly will first convert the market data of BTC-USDC of Coinbase into RGB images, and then send these images into thousands of simulated sensory inputs corresponding to the visual neurons of fruit flies. The system subsequently makes trading decisions based on neural activity.

In terms of feedback design, if there is a combined profit, the system stimulates 15 dopamine neurons marked as PAM11; if there is a loss, then it activates 2 neurons marked as PPL101 that are related to aversion signals. The experiment attempts to use a reward and punishment mechanism to make the simulated system develop trading responses.

However, at this stage, it cannot be concluded that it already possesses effective learning capabilities. The report mentions that this $1 profit may also be the result of short-term market fluctuations and random trading combined.

Drosophila connectome experiment dissemination

There has been a clear increase in such experiments recently, against the backdrop of new progress announced in the field of neuroscience earlier this month. Google Research and several other institutions have completed the mapping of the complete connectome of the adult male fruit fly's brain and central nervous system with the help of AI.

Public information shows that this connection map contains 166,700 neurons, covering the brain, visual cortex, and ventral nerve cord. With the release of this model, developers have begun to integrate similar simulation systems into more digital scenarios.

Wormuth has previously worked on a project named DOOMFLY, which involved integrating the same MaleCNS model into the classic game Doom. In addition, other developers have used fruit fly connectome simulations in projects such as Beat Saber, Super Mario 64, Minecraft, and Pong.

Despite the endless emergence of related demonstrations, researchers have not yet reconstructed a complete living brain. More complex processes such as neurotransmitters, chemical regulation, and gene expression still cannot be fully replicated in current simulations.