Nice summary. A long time ago I was offered a job to specialise in first generation AI. I saw an example of a mistake made by training a camera to recognise Russian tanks. It made big mistakes but the data was very small. Big data is the solution to this. The other example was choosing the correct material to make the inside of aero engines out of. My conclusion was that the experts were falling over themselves to make that decision. They enjoyed it. An AI prog may be able to suggest a marginally better solution but the data is so small (there are only a few dozen metals and ceramics to choose from). Suggesting a totally different approach to aero engine design is where the big money is.... and there is no data ... but there are some very clever people... I didn't take the job in AI .... I would this year, but as Terence explained, choosing the right place to use AI, isnt a job that AI would be very good at.... and humans are not very good at that either.