notes-cog-ai-aiAsAdvancedComputerScience

Awhile ago i heard that historically, there were some research topics that used to be called 'A.I.' but that became called something else after they succeeded. Some people are of the opinion that the field of A.I. research has been more successful than it seems because of this phenomenon.

In some cases, ideas from cognitive studies and A.I. appear to have cross-fertilized other fields, e.g. semantic networks and slot-and-filler structures may have been part of the inspiration for hypermedia and for object-oriented programming, although surely these had other influence as well. Also, Lisp was invented in an A.I. lab.

I've also noticed that the field of A.I., which used to be about creating general-purpose human-equivalent intelligences, now includes many more limited goals, such as systems for quantitative predication and approximate optimization (machine learning/data science); to the extent that the orginal goal has had to be renamed, with new names like GAI (General Artificial Intelligence) and Cognitive Architectures.

This leads me to reconceptualize much of what i was taught in "A.I." classes in school as "nascent advanced computer science", e.g. research looking, not for better ways of doing what computers can already do, but for ways to get computers to do things they can't really do (feasibly) yet.

If much of 'A.I.' is really 'nascent advanced computer science', then i figure if i'm interested in cognitive studies, i should also try to learn other 'nascent advanced computer science' areas, such as advanced programming language ideas.

Is there any part of A.I. which is not just 'nascent advanced computer science'? I postulate, yes:

The common ground between both of those criteria is that "true" A.I. research moves away from the paradigm of telling the computer, "here's a problem, solve it", or "here's a task, complete it", and towards the paradigm of "autonomous behavior".

This is tough because we are making it harder for us to have objective criteria for success; which makes results harder to publish, compare, etc.