If I was really really good, I'd advance the state-of-the-art in neural network-based machine translation, recurrent neural network language models, neural Turi

If I was really really good, I'd advance the state-of-the-art in neural network-based machine translation, recurrent neural network language models, neural Turing machines, etc. Right now, the state-of-the-art is exciting but falls short in some world-changing areas like question-answering. As much as I want to, though, I have no idea how to push the field forward, however. For the other things I want to do, I am probably sufficiently skilled to do, but I just need the time and motivation.