Good article in ComputerWorld on the current state of AI (Artificial Intelligence) . This time no mention that we will soon create intelligence as there was in the 80s and 60s. Emphasis on faster computing, more data from sensors, learning algorithms, dealing with uncertainty, agent models and integration of multiple disciplines. Coverage of examples of both research and applications. All of these applications are what we can describe as 'deeply embedded', not the high level, aggregation of business rules of the expert systems of the 80s. There are still expert systems style methods around, such as LPA, which I have covered here before.
As an Microsoft researcher states, these systems worked in narrow areas and were described as 'fragile' when they got large. We wrote a number of them and they ended up being difficult to build and maintain and would break when the knowledge had to be changed. Still, some were very profitable and lasted twenty years.
The article:
" Future Watch: A.I. comes of age
After decades of limited application, artificial intelligence is everywhere. And it really works this time.... " "
No comments:
Post a Comment