There is no such thing as artificial intelligence
(without some way to tell)

When the outcome is already defined It is an application.

Is 5 greater then 4 ?

Computers can determine new calculable facts, but they cannot determine new knowledge. The determination of what to do with facts must already be programmed, otherwise it has no value.

"AI" is a great marketing term, I guess much like "Emu" was a few years back.

But if you want to classify a program as "Artificial Intelligence"  then there should be some rule to apply, otherwise it means nothing.

  • Muscle Memory  knowledge that is taught (5 is greater then 4 )
  • Artificial Intelligence programs that always produce questionable results that are good enough for human acceptance. (Is a cat in this picture?)
  • Calculation, programs that always produce expected results
  • Simulation, programs that always produce questionable results

Marketing will of course classify every program as "Artificial Intelligence" . The classification of "AI" should probably be limited to the five human senses; sight, hearing, taste, smell, touch. If the program does not guess at one of the five sense, then you should probably assume the program is one of the other classifications.  If the program cannot be trusted to the level of a human doing the same job, then you should also choose another classification.

There are very few programs (if any) that can be classified as "AI".  We generally write "Simulations" then try and find a very specific use for the results. In the future we may get to some general use "AI" programs. I guess it's fun to play with the idea that "AI" is possible. Take the most worked on application "Character Recognition"  programs, they are very good, but do I use them? Only when they help. Would I classify them as "AI", probably not. We use this as a security filter, so we know the limits, and how to break them. But is this "AI" or just very specific use "simulations of vision".