AI was always a probability problem.. If we look at the emergence of sequence models (or statistical learning in broader sense), they predict the next sequence based on the accumulation of knowledge that humankind has acquired over the years.. The scientists who were tackling this problem(creating general AI) before thought of approaching it by creating different simulations for all kinds of problems, which would have led to infinity anyway, that is why it never worked. The solution was simple. Wdyt?
Comments
PaulHoule•33m ago
This was my opinion circa 2011 or so when I was recapping the old AI.
Consider, for instance, the successful early medical diagnosis program MYCIN
which like any kind of diagnosis process is a problem of reasoning with probability. Language understanding has the same issue, like if you wrote a grammar for English you'd find that common sentences have 1000s of possible ways to parse and you will need to either make a guess or keep your options open.
MYCIN had a half-baked approach to reasoning about uncertainty that worked, one of the reasons why symbolic AI fell out of favor was that nobody developed a generally useful approach to bolt probabilities onto logic.
jqpabc123•17m ago
Current AI is all based on probability --- and that is a problem.
PaulHoule•33m ago
Consider, for instance, the successful early medical diagnosis program MYCIN
https://en.wikipedia.org/wiki/Mycin
which like any kind of diagnosis process is a problem of reasoning with probability. Language understanding has the same issue, like if you wrote a grammar for English you'd find that common sentences have 1000s of possible ways to parse and you will need to either make a guess or keep your options open.
MYCIN had a half-baked approach to reasoning about uncertainty that worked, one of the reasons why symbolic AI fell out of favor was that nobody developed a generally useful approach to bolt probabilities onto logic.