4

I am currently considering using some Expert System for my application.

I really like the main idea behind the expert systems but I have no direct experiences with them. While I was searching through the internet I realized that the expert systems have become significantly less popular than in the past (no new books, outdate software, ...).

What is the reason for such decrease in popularity? Is its main idea overcome by some new and better approaches? Is it worthy to spend a time by developing some expert systems or is it better to try to find more promising scheme?

I understand that these questions are difficult to answer in general but I am just trying to find out the reasons for this significant decay.

Update: There are 10 main categories of expert system applications (Interpretation, Prediction, Diagnosis, Design, Planning, Monitoring, Debugging, Repair, Instruction and Control). My main question is weather there are some better alternative concept (for these kinds of tasks) which overcome the original expert systems.

I am mainly interested in the Design category (= Configuring objects under constraints by using many simple expert knowledge).

Czenek
  • 85
  • 6
  • 3
    Read about the [AI winter](http://en.wikipedia.org/wiki/AI_winter), and read also the [blog of J.Pitrat](http://bootstrappingartificialintelligence.fr/WordPress3/) which might suggest some explanations. You might consider that [business rule engines](http://en.wikipedia.org/wiki/Business_rules_engine) are some kind of expert systems today. – Basile Starynkevitch Apr 30 '15 at 20:19
  • BTW, you don't explain what kind of applications are you designing & implementing. So nobody could tell you if an expert system approach is relevant. At least, edit your question to explain what do you want your application to do. – Basile Starynkevitch Apr 30 '15 at 20:23
  • 1
    The main differentiator is whether the rules used in an expert system are **automatically or semi-automatically generated, and if so, how, and what data are used to drive that rules-generation process**. Otherwise, expert systems are no different than ordinary software programming, where business logic is first outlined by business analysts and then lowered into the source code by software programmers. – rwong Apr 30 '15 at 21:50
  • 1
    Indeed, it is quite common that a "big data" project chooses to productize and deploy its statistical insights in the form of an expert system. – rwong Apr 30 '15 at 21:54

1 Answers1

2

I studied A.I. in the 70s, and I never got too excited about expert systems. It seemed like the hype exceeded the content, which seemed to boil down to "have lots of IF-THEN rules and have a system for firing them". To me, that seemed like getting awful specific about something before understanding it very well.

Nevertheless, some interesting expert systems were built. DEC had an expert system for configuring VAX computer systems. Supposedly there is an expert system currently being used for scheduling airline gates. I'm sure there are a lot of other examples.

If you look at current on-line search engines, and IBM's Deep Blue (or whatever it's called) those could be called expert systems. GPS systems and Siri are pretty impressive. At least they seem to know a lot, though they don't seem to do much in the way of thinking, understanding, reasoning, learning, etc.

It may be that, these days, there doesn't seem to be a lot of need for that kind of software, compared to more mundane stuff.

Mike Dunlavey
  • 40,059
  • 14
  • 91
  • 135
  • Thank you for your opinion. And as you said, there are several impressive applications of the expert system - and despite of this success its popularity still decreases. Wierd ... – Czenek Apr 30 '15 at 20:34