- I already wrote it above, and still believe it is true: Such problems, where the "actual goal" is to give cutting-edge research an order-of-magnitude advantage, is not suited to securing a network of transactions. I wouldn't want to run an exchange and accept a currency that could be 51%-attacked every time someone made a break-through.
This is kind of the difficulty-time-warp problem in a nutshell. This is why I'm thinking that the work function needs to also include a scale of computational complexity in addition to the challenge target. We should have two difficulties, one to throttle block production in general, computational complexity, and one to scale the difficulty of the challenge, maybe we could call this "cognitive complexity."
(HunterMinerCrafter stated himself above that he could 99% Motocoin if he wanted to. While trusting him not to do it is possible, I wouldn't want to have lots of money at stake with that. Then I could just put it into the bank, after all.) Having the currency "just for fun" and not intended/advertised as fully secure (as Motocoin's website still does) makes this point moot.
There was a time when btc had only a small handful of miners as well, any one of whom could have 99%ed bitcoin then, too. Every coin goes through such a phase. The beauty of a blockchain is that you don't necessarily have to just trust me implicitly, instead you could just fire up your own bot rigs and dilute my hash rate. (The difficulty-time-warp attack vector aside, anyway...) As more and more participants choose not to blindly trust each other the network eventually becomes strong enough that no one has to.
- While I'm all for promoting research (I'm on a research grant myself, although not related to AI or crypto-currencies but applied mathematics), I also think that results should benefit the community at large (keyword "Open Access publications", for instance). I doubt that the research done for bots on Motocoin or a similar currency would at all be published somehow, as it is in the botters' best interests to keep their results in secret. Why should the community of currency investors pay them for something they research just for their own profit?
They probably wouldn't, directly... at least not while they are still "cutting edge" in terms of coin production efficiency. However I'm sure that once the naive bots are no longer productive they would be released for study, once there's no longer direct incentive to "keep the secrets."
Also advances in algorithms and infrastructure that arise out of work on moto bots could easily find their way into other systems/products, and benefit everyone indirectly.