The arc of progress in AI is pointed decidedly _towards_ explanations, not away from them.

I don't dispute that people such as Anderson and Weinberger were promoting this instrumental view of AI, but the idea didn't survive scrutiny. (Norvig called out Anderson explicitly for misrepresenting the case back in the day.)

It's a category mistake to confuse observations with understanding (even if those observations are produced through DL). People wish there was some magical process for creating knowledge, but unfortunately it remains hard (and human) work.

Thanks for highlighting your article to my attention.