AI is to computer science what black magic is to science.

Seriously, what do you get after you’ve spent days and days to train a model? An inscrutable blob that may as well be proprietary software written for an alien CPU; studying it is damn near impossible, understanding how it works would require several lifespans, and yet it works, and we trust these models and use them to get solutions to problems that would normally be impossible to handle by computers using “real” computer science. And one day, this trust will bite us in the ass, not in the form of an “AI rebellion” but with every system that uses AI becoming unreliable because of situations outside its training.