Quantum Blockchain Technologies (QBT), a company dedicated to the research and development of “Blockchain” technologies, published on June 19 a statement in which he claims to have optimized the operation of the Bitcoin ASICs through artificial intelligence (AI).
Development introduces automatic learning models, with experimental support, which would optimize the search for Nancethe random number that miners must find through millions of combinations, in order to register a block of transactions in Bitcoin accounting and obtain rewards.
QBT proposed three different approaches, called Method A, Method B and Method C, to predict unances with greater probability of success, potentially improving process efficiency when requiring a shorter energy consumption time.
Method A, according to the Report From QBT, it is a development-based development that aims to reduce the search space of the SHA-256 algorithm compared to the traditional method.
As explained, QBT compared two identical Asic miners, operating in real time: a team with CGMINERopen source software used for GPU -based miners (graphic processing unit), FPGA (Arrangement of programmable logic doors in the field) and ASIC; Then, another ASIC with a modified version of CGMINER that included QBT AI software.
Although the QBT statement did not contribute tests or technical details Relative to this comparative procedure (such as the processed blocks), the results presented by the company reflect that the miner with the QBT software generated greater rewards per unit of time, evidencing an increase in the quality of the hashrate.
It is important to clarify that it was not possible to verify those results in publications In social networks or ads Prior to the company.
On the other hand, method B combined automatic learning techniques and statistical optimization to address the same problem from a different approach, although complementary to Method A.
The QBT team said that this method also significantly reduces the search space of the Nancewhat could translate into lower energy demand and greater profitability for miners.
The C method, on the other hand, represents a more ambitious advance. This approach combines an AI model in training with an “oracle of AI”, a system that evaluates in real time the probability that a Nance Generate a winning hash.
Unlike methods A and B, which are integrated into mining software, Method C requires modifications in the design of ASIC chips, which would make it more appropriate for new developmental hardware models.
As said by the company, QBT He is already collaborating with chips manufacturers ASIC to evaluate its integration.
Implementation and market challenges
To implement these proposals, QBT focuses on secondary market control plateswhich replace the originals of the mining equipment, especially in those made by Chinese companies.
Those plates, that They optimize hashrate and energy consumption through techniques such as Overclocking or voltage reduction, are a key segment of the market.
QBT plans to integrate its modifications into two operating systems: CGMINER, described above, and Espmineran open source operating system based on microcontroller ESP32specifically designed for Bitcoin mining devices.
According to QBT, “the implementation of the modifications of CGMINER and ESPMINER in secondary market control plates does not represent a technical barrier”, since the suppliers of these plates make similar updates routinely.
Francesco Gardin, CEO of QBT, said: “The demonstrations in Las Vegas (where the Bitcoin Conference event was held) were essential to convince the main actors that the ability to predict the behavior of SHA-256 with AI learning models is real, since it offers a practical solution to improve the quality of the hash’s power of a miner.”
The QBT approach not only seeks to optimize the performance of the ASIC, but also guarantee continuous control over their technologies. The company plans Maintain a live connection with your servers To manage the use of their AI models, which could imply a subscription or licensing model for miners.
Although the open source nature of Espminer allows community audits, QBT owners could lead to Dependency of that live connection with QBT servers. This could introduce a risk in centralization if this technology is adopted massively, compromising the autonomy of the miners.
Finally, despite the statement that these models can increase the performance of Bitcoin mining, the need to validate it with real figures arises, because the extra energy spending involved in using AI may not be compensated with the benefits of getting blocks more quickly.
