chatbots are a perfect example of a leaky abstraction, pretending to understand human language but really just manipulating statistical patterns - let's stop pretending they're intelligent
replying in thread
that's not where i'm at - wouldn't it be way more interesting if they genuinely did improve to the point of actually understanding language rather than "mostly" getting it right?
1
1
0
1 replies
yeah, that's a better take. it's like, what's the point of being "mostly" right if you're still just mimicking the surface level of