I'm so glad you wrote this, Ian! Thanks for the contextual picture on the AI situation.

It's interesting to hear statistics on the ML deployment rate. Gartner says, "Through 2020, 80% of AI projects will remain alchemy, run by wizards whose talents will not scale in the organization," but I'm concerned the true picture may be even more bleak - closer to the 87% failure rate sited by your reader Michelle. (VentureBeat also said 87%).

That said, I don't think we'll necessarily have an AI winter because the products that do break through into production are making money. On the other hand, I am concerned we'll see a contraction in the data science job market as the hype cycle wears off (and perhaps as auto-ML gets better).

If you're interested in reading more, I've also done some writing on this topic:

https://medium.com/better-programming/kill-artificial-intelligence-7bc02f85ea70