This is a point worth emphasizing. So many proposed "AI" use cases insist on trying to make some kind of end-to-end prediction, rather than let ML do something it is well positioned to do (e.g.) search more effectively, and let a person make the judgement for the things that aren't so readily predicted.

My analogy is a weather forecast. Regular forecasts give you the temperature and precipitation, and let you decide what to do based on this. If meteorologists instead black boxed the weather and just tried to tell people what kind of clothes to wear every day, I suspect people would start ignoring the forecast. But this is how so many ML use cases are set up.

I'm happy to see an application that respects ML for what it can do and not try and shoehorn it into the "prediction machines" paradigm.