Here's something new. Not sure how artificial intelligence is involved, but a guess is that it's a deep neural net designed to recover spoken words in the presence of noise and then articulate those words out in a separate channel with the noise gone.