There's a big difference in this project which makes it more interesting than others. Well, at least from what i've read. AI projects i've seen are focused on such things as creating distributed AI themselves, or creating distributed app store, or data mining from other people phones, etc, and monetizing everything. On the one hand, it's sound a little utopic, and on the other hand, monetizing can be a fail idea. In this industry there is a lot of big and comperhensive datasets for many tasks (ImgeNet, MS-COCO, etc), which are free and constantly improving by community, many new specific datasets created using these ones. A broader datasets is one of the thinks that are in big demand to create more sophisticated, universal AI. Monetizing it wouldn't be popular. The same with source code for many algos that available on github to improve and use it in own work (no plagiarsm, ofc). Decentralized AI - well, you just invest in AI algo developed by startup. The success of it is questionable, and it's limits, the qualification of team and their goals in the future...

In Myriads.IO, on the other hand, there's no AI development or monetizing something that already open-source, free, and developing with rocket speed. They propose exactly the thing needed for AI research in general and scientists in particular. GPU power. Miners want to sell a lot of it, AI researches wants a lot of it, cheaper*. And not just miners, AI scientists have GPUs too, it might be handy to rent it 24/7, and use the platform when you actually have some experiment to conduct, without touching the "farm".

*_{I can confirm that it's often just too expensive to use current cloud GPUs for experiments, as said in the video. I don't have enough GPU power to train my nets efficiently,
 but i'm not ready to pay for amazon or nvidia gpu cloud, it's just expensive as hell (much more than you can theoretically mine with such GPU power), i'll better wait for my local GPU. There's a big gap between young AI enhusiasts that don't get paid much for their science interest and big companies which can pay for these servers easily}