You do wonder where it will end up.
It's certainly been a long time coming. 30 years ago I was programming neural networks in the first generation of so-called 'Expert Systems', which despite the hype were little more than complicated sieves to figure out which inputs could be handled automatically and which needed a human expert to look at them.
We were doing it for underwriting life insurance cases where humans have the good grace to die by and large in accordance with the established actuarial tables, but there was loads going on in other branches where the risks weren't quite as well understood.
