The ability of a machine to understand a simple sentence like a horse riding on top of an astronaut, has not really improved. The systematicity problem has been known for forty years, and there has been no real progress on it. It is still an unsolved problem. And the intellectually honest thing to say, i think, would be, hey, we've done a great job with a certain sub set of problems, but artificial intelligence is hard. We don't have the eye that can do anything like that.