Reigniting The Computational Arms Race

How can neuroscience research help fund research into new AI chips?

Bitcoin’s initial pitch was as a network where donated computing power would lead to the end of botnets, spam, and other malfeasance the internet had spawned. The idea was that instead of using their computing power for nefarious means, people with large amounts of underutilized computing power would be able to redirect their processors towards securing the bitcoin network, in return for a tangible financial reward.

As bitcoin mining became more profitable, what ended up happening was more akin to a computational arms race. Graphical processing units (GPUs) quickly became the weapon of choice over traditional CPUs in the ever-prospering bitcoin mining industry. Because GPUs are used by researchers in medicine, aerospace, and practically any field of science where there is large troves of data (i.e. everything), there was a brief moment in bitcoin’s history where it was doing tremendous good for the world. Research and development in faster processing cards for bitcoin mining was trickling down substantial benefits for poorly funded areas of clinical and other research.

That all changed with the advent of ASICs – dedicated hardware for bitcoin mining. Because bitcoin’s mining algorithm is relatively simple, and does not rely inherently on machine learning, as a single bitcoin became worth tens to hundreds of dollars, wise investors started developing power-efficient computer chips that could be used only for bitcoin mining. Bitcoin’s arms race had reach its full potential, but it was no longer doing good for the world outside of bitcoin. Sadly, a mining market that had once shown potential for improving nearly every big data problem faced by mankind, was now suffocated by the flow of capitalism.

Because Project Oblio is a network that rewards people for correct answers to a machine learning task, it is possible for us to re-ignite the computational arms race first started by bitcoin. Early on, Project Oblio will likely run on the same GPUs  that once propped-up bitcoin mining. However, as the network becomes more useful and thus valuable, we can expect custom-made “neural network hardware” to reach the network due to its increased profitability in mining oblios. These same cards can also be used in more general machine learning tasks, performing much better than GPUs, when conducting analyses in the less-well-funded scientific  areas described earlier.

Thus, an investment in Project Oblio is a long-term investment in big data technology. When you support the network, you are supporting all areas of science that analyze big data – cancer research, population health, self-driving cars, etc. A computational arms race that drives machine learning hardware research is something few other systems like Project Oblio can offer.

Leave a Reply

Your email address will not be published. Required fields are marked *