I think the term you’re looking for is “willfully ignorant” or perhaps “stubbornly ignorant”. They’re actually too fundamentally intelligent to be truly stupid, but the intelligence is being perverted in the service of an active ignorance that seeks reinforcement of its own preconceptions, rather than being open to the subtle panic that inevitably arises when learning something truly new & different.
I came across this paper (actually a literate haskell document, pretty cool!) from MIRI. (What’s MIRI ? I suppose someone finally convinced the “Singularity Institute” they had a terrible name) So I’ve come up with a worthy side project for a while: coming up with an interactive counterpart of the same written in Common Lisp. It won’t be easy, since (1) I’m sort of a noob right now, and (2) this was created by a couple of people working full-time for a month, but it should be fun and rewarding. Since they call it BotWorld, I can call my version BotVerse 🙂
Update: I tried this, but translating haskell to lisp is no fun … also, I don’t like some parts of how this is implemented, and I’ll ping back here when I have something to show!
It has been hypothesized that whatever intelligence is (and we obviously have not been able to fully define it so far), it is a multidimensional thing. Natural human intelligence appears to be multifaceted.
“Human-level intelligence” does not lie at the end of the road that focuses only on machine learning algorithms, nor will it be the result of the world’s greatest natural language processing machinery taken in isolation. It is somehow a combination of these things and more.