... if they aren't doing the learning process, then this would be a fairly useless project.

They provide little information, this is my conclusion after reading whitepaper. And I also struggle to find any usefulness, if that's the case. On the other hand learning process requires high memory bandwidth, powerful GPUs - distributed computing seems not to be compatible with this, so would be interesting if they have an idea how to solve it. But it would require new NN architecture / learning process, so I'm sceptical...