Not Ready for Real Neural Nets

I'm working on the Human Brain Project, and in particular doing a lot with Neuromorphic computing.
We don't really understand how the brain works, but I think we're making good progress.
One thing that might be useful for you all is the parallelism we can get on say the spinnaker platform. We can simulate 500,000,000 point model neurons in real time.
Unfortunately, we don't really know how to take advantage of that.
We're working on robotic applications, as that may be a good use.
I'm hoping to get to large associative memories, but that's not ready yet.
I'm also interested in big cognitive agents working in virtual environments. We're not really ready to learn about those environments yet.
I'm pretty confident I could do a really vast finite state automata with lots of states simultaneously active more or less now. If you have an application where that would be useful, let's talk.
Otherwise, I don't think this stuff is immediately useful for solving current problems without a lot of work.