Re:Machine learning works spectacularly well, but mathematic...
by jldugger on Friday December 04, 2015 @08:19PM
(#10680350)
And for the first fifty years of AI research, the process more or less goes:
1) An AI researcher decides that problem X is hard enough that solving it represents intelligence
2) Researcher develops an algorithm to solve X
3) Now that we have that algorithm, it's no longer an AI problem
When you teach a computer to play chess, it's just depth first search with a few heuristics added in. When you write an expert system to evaluate mortgage loans, it's just following a if-then script.
Machine learning techniques though, seem to to have stayed solidly in the AI camp.
1) An AI researcher decides that problem X is hard enough that solving it represents intelligence 2) Researcher develops an algorithm to solve X 3) Now that we have that algorithm, it's no longer an AI problem
When you teach a computer to play chess, it's just depth first search with a few heuristics added in. When you write an expert system to evaluate mortgage loans, it's just following a if-then script.
Machine learning techniques though, seem to to have stayed solidly in the AI camp.