To be fair to those assumptions, there've been a lot of cases of machine-learning (among other tech) looking very promising, and advancing so quickly that a huge revolution seems imminent—then stalling out at a local maximum for a really long time.