Quantum Blockchain Technologies PLC (AIM:QBT, FRA:BYA1) said its Method C AI Oracle produced a consistent advantage over traditional Bitcoin mining in tests conducted after the software was adapted for an ASIC manufacturer’s mining rig.
The company’s researchers have completed the reconfiguration of the Oracle’s learning models for the rig’s operating system and rolling architecture. QBT said the architecture differed materially from the Bitaxe Gamma platform on which the technology was originally developed, requiring the models to be rebuilt around the ASIC rig’s distinct data characteristics rather than simply retrained.
An initial model was presented to the unnamed manufacturer on 28 July, with a further update scheduled in the coming weeks as QBT gathers additional data and works towards the performance previously achieved on the Bitaxe Gamma. The company did not disclose a quantified improvement over conventional mining.
QBT is also compressing and optimising the software so it can run directly on the manufacturer’s hardware. Testing of a compressed version has already produced a significant increase in processing speed, it said.
Chief executive and executive chairman Francesco Gardin said the results must remain robust over longer periods before the company is ready for live testing or can produce a commercially viable product. QBT’s next steps include broader training datasets, further model optimisation and a live demonstration to the ASIC manufacturer.
