The Student Room Group
Reply 1
It depends how you define AI, a lot of modern data mining techniques were developed under "AI research" but aren't really about approximating human intelligence. NN has been seriously stuck in the mud for decades, ever since Minsky smacked the field around with his book "Perceptrons" little real progress has been made.

Some notable practical achievement have occured (The computer backgammon player to be undisputabley better than humans was based upon neural networks), but little in the way of theoretical progress.

One problem with AI is that as soon as an "AI" problem is solved it stops being an AI problem. OCR used to be considered an AI problem until Abbyy Finereader showed it could be done just fine using some clever heurisitcs.

We're still a long way off from mimicing everything a human can do though. In part this is because there are so few good computer science researchers as the best comp sci students have so many options available to them that few chose to go into academic research.

Latest

Trending

Trending