Preaching generative AI in 2019 is like telling everyone how great Radiohead is in 2009. The writing was on the wall; Google BERT came out a year earlier, Tensorflow had been part of computer-vision pipelines for almost a half-decade and GANs were "so last-year". In a lot of ways the industry was <i>already</i> going through a practical AI transformation, driven by workflows that leveraged autonomous feedback loops.<p>Compared to the industrial and military adoption of AI, 2019 is comically late to the party. The idea of a machine generating text was about as "wildly ambitious" as a knife cutting bread. Imagine the wonder on their faces if you demoed a Markov chain to that crowd.