It may come as a surprise to some, but it is possible to implement the Bitcoin protocol simply by using phone calls, paper and pencils. This is obviously not the best way to go about it, but it does serve to emphasize that Bitcoin is not inherently digital, but a digitally implemented concept. A concept that is not without its problems. Yet to find creative solutions to a challenge, one must sometimes look beyond the obvious, to explore what is overlooked.
Bitcoin has always had a bad reputation for being a poorly implemented and power-hungry technology, due to its consensus mechanism, proof-of-work. A criticism that is only half deserved, since proof-of-work has proven to be a simple and reliable system, providing a backbone to a crypto-currency market rich in innovation. And as such, enabling the existence of other financial systems that themselves emit almost no energy but have yet to prove that they are on the same level of reliability as their predecessor.
As a reminder, proof-of-work is a consensus mechanism that allows the world to agree on changes to the same transaction ledger. To do this, participants compete to generate random hashes until they find one that matches the rule imposed by the protocol. The person who successfully generates the hash wins the privilege of submitting the next change to the ledger.
Despite the use of application-specific integrated circuits, the best hardware today has a power consumption of 3250W to generate 110 Trillion hashes per second (Th/s). The proof-of-work model thus consumes a lot of energy. And this is mainly due to the fact that the generation of a hash requires to perform a series of operations, those that are today carried out digitally, i.e. on bits of 0 and 1.
The use of the digital medium to perform operations is interesting because it is very versatile, and properly programmed, it can calculate anything. Its antonym, the analog medium, does not use bits to make calculations, but a continuously variable physical quantity such as an electric current. And it is the physical conditions in which the current passes that will modify it and allow a solution to the operations to emerge. By definition, the analog model is tied to the physical layout of each piece of equipment, which is therefore only capable of providing what it was built for. From this lack of versatility also comes an advantage for analog, its use of infinitely less energy.
Current research in terms of analog processors is still in its infancy. The field of machine learning is at the forefront of this race to break out of the all-digital paradigm as it is a field reliant on performing computationally expensive operations to train models over days or weeks. But the incentive for bitcoin miners is much the same as that of AI researchers, so it wouldn't be surprising if a trend toward crypto-currency analogy emerged.
In this quest, two schools are likely to emerge, those that will attempt to adapt Bitcoin to existing analog technologies, and those that will attempt to adapt analog technologies to the rules of the Bitcoin protocol. This latter, has yet to be realized, as none of the hashing algorithms or digital signatures used by Bitcoin have been researched in the analog domain.
However, attempts to use existing technologies to create an efficient proof-of-work have already been made and, although they have never gone beyond the prototype stage, they have paved the way for further research on the subject. These approaches have focused on finding challenges similar to those of current hash generation, but which could be applied to analog processors.
Per example, throughout the past two decades, silicon photonics has experienced rapid growth and improvement that led to the commercialization of silicon photonic coprocessors. These analog integrated circuits make use of photons instead of electrons to perform efficiently specialized computing tasks, making them good candidates for mining a cryptocurrency using an optically computable hash.
Analog Hamiltonian optimizers have also emerged as good candidates for this task. This type of "analog quantum simulation" develops simpler models of complex quantum systems while reproducing all the physical attributes of interest, allowing them to solve very difficult computational problems very quickly and efficiently.
In terms of performance, claims range up to 85% reduction in power consumption through the use of photonic processors. Very good news to be taken with a pinch of salt due to the lack of independently verified tests in this area.
Finally, despite the existing possibilities in analog crypto-currency innovation, the competition for more efficient consensus systems is bringing more developments every day and is probably more likely to be the ultimate solution to the energy consumption problem.