FutureBit has unveiled HashFly, an unusual proof-of-concept that uses simulated neural activity from a fruit fly connectome to perform simplified Bitcoin hashing. It is nowhere near replacing modern ASIC miners, but the experiment raises an interesting question about what ultra-efficient computing might eventually look like.
Bitcoin mining hardware has spent years moving in one predictable direction: smaller chips, higher hashrate and fewer joules consumed for every terahash.
FutureBit has now tested something rather less conventional.
The company behind the Apollo mining hardware line introduced HashFly, a browser-based experiment that uses 2,914 simulated neural traces from a digital fruit fly nervous system to perform simplified calculations related to Bitcoin's SHA-256 hashing process.
The experiment became possible after Google Research and its collaborators released a detailed connectome of the adult male fruit fly's brain and central nervous system earlier this month. The MaleCNS v1.0 reconstruction contains roughly 165,122 neurons, giving developers a remarkably detailed digital representation of biological neural circuitry to experiment with.
For now, however, HashFly is exactly what its creators describe it as: a proof of concept.
It does not mine at anything close to real Bitcoin network difficulty, and its performance is tiny compared with dedicated silicon. Tom's Hardware observed roughly 100 kH/s in the browser demonstration, while FutureBit's compact Apollo III is advertised at up to 18 TH/s.
That enormous gap is why the experiment is more interesting as an energy-efficiency study than as mining hardware.
FutureBit says that a hypothetical system built with real organic neurons could, if the concept scaled successfully, reach around 1 watt per terahash. The company describes that figure as an estimate based on the fly's total biological energy use and the assumption that all available neurons could continuously contribute to hashing. It has not been demonstrated by a physical Bitcoin miner.
FutureBit says it plans to extend the experiment across more of the available neural dataset and publish its findings as it explores how SHA-256 could be mapped onto the larger system.
There is little reason for ASIC operators to worry about biological miners appearing in server racks any time soon. Modern ASICs remain vastly more practical, predictable and powerful.
Still, HashFly is an interesting reminder that mining efficiency is fundamentally a computing problem. Today's competition is measured in joules per terahash. The technologies used to improve that number decades from now may not necessarily look like the chips inside today's miners.
