Not really interested, but if you have cycles to spare on a GPU I may pay you (or the etherium network which you can join) to run my deep learning training. The demand for this sort of processing is only going up.
Why would you, though? As far as I understand this, anything calculated on the etherium network is essentially calculated by all miners. So it must be much more expensive than just buying the computing power from Amazon, otherwise the miners would not do it. Running something on the network will always be more expensive than getting computing power of your own.
What it offers are redundancy (cannot be taken down without involving the entire network) and independence from a central entity. This is interesting, but only really extends to contracts within the network. As soon as you want deals related to things outside the network, you need some way to get the data into the network and I currently do not see how to do that without reintroducing central entities.