Ted: This is why I do not believe that the “Turing Test” is really a test for intelligence, it is merely a best approximation because we are unwilling to accept the breadth of the concept we want to test for.

That was excellent. Deserves to go in a book somewhere imo.

Ted: we need to accept that they each may provide a tool or modular solution to different aspects of general intelligence.

One thing I believe is that the tools are unimportant. Its what can be done with it. You can design a system without a moments thought as to what tools will eventually be used. All that is needed is an appreciation in general of the tools available. To that end, I'd agree that ANNs, GAs, etc don't neccesarily mean AI, they're just part of the toolbox.

I don't think people pay enough attention what is meant by common-sense in an AI context. It's just a side effect, and end product. I'm sure I've met people without common sense. I've even been lacking it sometimes. But if by common-sense here you mean advanced knowledge of the environment that we take for granted, being able to look around a room, identify three-dimensional objects in the environment, remember what they are, what you have used them for, etc, then all common-sense is, is goal related.

For example, make a tribal hunter from the african plains close his eyes, then when he opens them he finds himself in a comms room... I don't think he is going to display any kind of common sense you or I would recognise. Why would he? He's never needed to before. All his common sense is only relevant where he grew up, where he lives. Its not needed here. And vice versa, I'd say. Also, when we are born we have no common-sense. Its something we gain over time, an accumulation of all we have needed to learn about the things in our environment.

If AI isn't being developed with common-sense today, I'd say its because AI projects under development aren't trying to achieve the right kinds of goals that would lead to this side effect of common-sense.

Can common sense in one system be downloaded into another?

If following a computational metaphor (way too closely), then it would be a case of both systems have similar enough data structures, data storage, or data languages.

But how do we articulate this same concept NOT knowing how these systems work?

I don't think we should, and I think this is where people conceptualise this particular form of common-sense wrongly. You cannot expect a system displaying common-sense to just download it to another system, or do a file dump.

How long does it take us to teach something to someone? How long does it take to learn a subject? Years.

I would say that transferring common sense from one system to another would only be possible if both systems had similar goals.

And further, it could take years*

:)

Unless someone comes up with some new good ideas.

(*All a worst case scenario of course. To do properly of course, depending on how complex and detailed this common-sense information is).

Gregg

Also, another thought just occured - Perhaps it is not the transferrence of common-sense knowledge that takes years - rather bringing a system up to scratch into a sufficiently advanced mode is what would take years, then the required sophistication for transferring and translating common-sense info would be present.

Its a little abstract I know, but in the absense of actual system workings I thought it okay.