Carlos E. Perez has put up a Medium essay which has a link to a DARPA video (16 minutes long) with a very good summary of where Artificial Intelligence is now and where it needs to go next, from their point of view.
As Perez notes, there are many approaches to AI currently being explored that the video doesn’t mention, but it does tell you a lot about the subject, and in a pretty easy-to-understand form. He gives some links to more online material.
Neither the video nor Perez’s essay mentions machine translation specifically, but they do give us some background for understanding how AI and MT relate. I would say that the future of AI research, as presented in the DARPA video, may go some way toward improving MT, but the big gap between machine and human translation would still exist even with this “third wave” or future wave of AI that the video describes. That gap is that natural languages, like English and Japanese, are vastly complicated human activities, and computers are still extremely far from the “Commander Data” stage of development, in which they behave fully (or pretty fully) as humans.
“Pattern recognition” and machine speech capabilities as they currently work may be somewhere vaguely in the area of the semantic and pragmatic aspects of natural language, but who knows if and when they will really reach those aspects, which are at the heart of translation in the true sense.