And This Is A Really Important Issue.
Normally I wouldn’t bother with the speech recognition/translation aspects of Google amazingly good, solid translations apps, but the final lines of the article intrigued me, because it’s so obvious than AI narrow platforms are headed this way, and it’s even more obvious with AGI. And this is a REALLY IMPORTANT ISSUE with ethical guidelines, and I think is going to bite us in the ass bigtime in the DARPA (and counterparts throughout the world, top-secret all) military sphere — which they already are working on furiously, although they will deny it.
“But while machine translation keeps improving, it’s difficult to tell how neural networks are coming to their solutions, says Bahdanau. “It’s very hard to understand what’s happening inside.” Keep it in mind.”