Since computers are basically just very complex calculators and it is very difficult to define complex interpretation and evaluation tasks in terms of  calculations, interpretation and evaluation is one of strong AI's biggest challenges.

Create expert systems or knowledge databases or knowledge management systems or whatever and have them perform in a sort of interactive relationship with human beings.  Automate what can be automated and let a human operator do the rest.

So now we have a kid riding a bicycle with training wheels.

Progressively, with time, you will take out the functions operated by the human being and you will have the machines do them.

If it is possible, all the better. You'll have a kid running the bike without the training wheels.

If not, you'll have an excellent KMS or interactive database system.

This is the way SAI, KMSs and Expert systems should be presented to possible investors: in a best and worse case scenario: at best we'll have a SAI program at worse a KMS or an expert system.
----
Could the author refactor this page to add more technical information or better descriptions? As it is, this page is nearly content-free (also an OrphanPage and, hence, a DeletionCandidate), although it does seem to advocate DoTheSimplestThingThatCouldPossiblyWork. . . but doesn't follow CodeUnitTestFirst.

----

"Just very complex calculators?" Consider JustIsaDangerousWord. Consider that the entire '''universe''' is apparently '''just''' a very complex calculator.

Every thing and process that has existed or might exist can be modelled by, or perhaps even '''is''', information and information analysis. "Calculation" means infinitely more than the 4 functions you find on a pocket caluclator.

On the other hand, we can comfortably conceive all things - in fact I just did it - all things. Maybe it is useful to categorize all things - like - "all things can be referenced as named collections of things."  And I am sure that there is a comprehensive set of the actions which all things perform. They all act, in some way. They all perform a function. Or hold other things, which implies actions like Get, Alter, Save, Delete, Select. I do not think the abstract set of operations is far from reality now. It is just these awful languages we have which makes progress in software difficult. I Tops me hat to ya. --PeterLynch

Also "interpretation" and "evaluation" are not well-founded here; they might mean something very narrow or they might mean something so overly-broad as to be meaningless.

The state of the art of AI and strong AI is that we've gotten to be quite good at finding ways for software to do extremely well those mental activities that humans are largely very poor at, such as playing chess, and we are still mostly stymied at getting software to do well at mental activities that humans find trivially easy, like recognizing all of the objects in their living room.

All of those things that humans find trivial are currently lumped together under the topic "common sense", so if you want to look at what is needed to accomplish strong AI, mostly it is about "calculating" common sense. But this has very little to do with addition and multiplication.

-- DougMerritt

-----
CategoryArtificialIntelligence