They are just statistical algorithms. Making good use of them requires demistyfying them, making them more transparent, validating them, having confidence tests

They are just statistical algorithms. Making good use of them requires demistyfying them, making them more transparent, validating them, having confidence tests and other indicators of how reliable any given result, and finally, human intelligence double and triple checking what the hell is going on. But that level of caution goes against the strategies people currently employ to draw attention, obtain funding or sell. So we have to sit back and endure the spectacle until logic reasserts itself. You can always fit a line to a cloud of points but using the result for anything important is a science in itself. This is very much the future of good ML/AI work.