This is a great question.
My remedial guess is that the human mind is more efficient at the pattern recognition that LLMs excel at in their own right.
We can do a lot more with less data, exert less effort and come to a reasonably accurate conclusion.
LLMs can artificially reason, but it requires intricate software that took decades to develop to the standard that it's reached now, and computers that suck the earth of its resources at a hair-raising scale, and like you've mentioned a lot of data. A lot of data. Apparently the entire internet and then some on a carousel.
Intelligence is an innate faculty of man and man's measure of intelligence generally doesn't require that much, depending on what's expected of the man throughout the course of his life.
Because AI is a technology the expectations we place on it are way higher.
A manuscript with a few errors, blotches, misspellings, omissions, what have you, is excused. If your printer does the same thing for every four or five jobs, it's defective.
My remedial guess is that the human mind is more efficient at the pattern recognition that LLMs excel at in their own right.
We can do a lot more with less data, exert less effort and come to a reasonably accurate conclusion.
LLMs can artificially reason, but it requires intricate software that took decades to develop to the standard that it's reached now, and computers that suck the earth of its resources at a hair-raising scale, and like you've mentioned a lot of data. A lot of data. Apparently the entire internet and then some on a carousel.
Intelligence is an innate faculty of man and man's measure of intelligence generally doesn't require that much, depending on what's expected of the man throughout the course of his life.
Because AI is a technology the expectations we place on it are way higher.
A manuscript with a few errors, blotches, misspellings, omissions, what have you, is excused. If your printer does the same thing for every four or five jobs, it's defective.