10 Learning Writ Large (2017-2023)

OK, imagine the year is now 2017, and neural networks are dominating almost all of AI. Language models (LMs) have made progress due to recurrent NNs. However, they still have problems keeping track of a topic for more than a sentence or two. Google is switching to recurrent NNs for machine translation (MT). Hold on to your hats, though—we are about to see an explosion of progress that has left me breathless.