This discussion floats between the two poles of AI - on the one pole is the idea that AI reduces to simple instructions ("it is not easy for a computer to grade an essay in anything more than a very functional rubric-informed way") and on the other pole is the idea that AI is simple pattern recognition ("it is designed to give you an output that looks plausible"). Both are caricatures. A grading AI doesn't actually use a rubric; it's a question of 'recognizing' based on ten thousand unnamed parameters) when an essay is an 'A' and when it's an 'F'. And at a certain point, with enough data (and the right data), the 'most plausible' response is almost certainly the 'right response'. But even more to the point - if people don't think that these are the sorts of processes that produce knowledge and intelligence, either in a human or an AI, then what do they think are the right sort of processes? Personally, I think all human intelligence is based on pattern recognition. I would need a very good argument or demonstration to convince me otherwise, especially after seeing what underpowered and data-impoverished pattern-recognition-based computing devices like GPT-4 can do.
Today: 0 Total: 273