The Turing Test

[geni]

Incidentally, kidding apart, does dyslexia generate a distinctive pattern of spelling error?

Sometimes

If so, how distinctive?

Would uoy say this is disrinctive? Reverseing words is fairly classic in some forms although that can be caused by other things. There is no single set of symptoms that all dyslexics display.

Distinctive of person? Or of type / degree of dyslexia? I suspect I could pick out a significant selection of Geni posts from a random bunch of other JREF posts by error pattern (if I could somehow program myself to ignore content, which would likely be far more effectively diagnostic). Pillory would be even easier, but would he pass a Turing test?

I wonder if everyone has a distinctive error pattern that a program would identify? (Most of mine are typos rather than conceptual verbal errors, so my pattern might be more keyboard dependant than personal).

Programs can certianly pick up writeing styles although it goes beyond error patterns.

 

[Underemployed]

The bad news is that you could probably avoid all that by uploading wikipedia into it. It would give it a knowage base that would rival pretty much any human. This has been looked into but isn't really posible yet.

I think we might be talking at cross-purposes here. A vast knowledge base of facts is not the same as the experience of existing in human form and interacting with humans on a personal level. I wouldn't consider Wikipedia as being any more self-aware than my local library. It's the application of what knowledge you have that makes you capable of passing a Turing test, not the knowledge itself.

 

[geni]

I think we might be talking at cross-purposes here. A vast knowledge base of facts is not the same as the experience of existing in human form and interacting with humans on a personal level. I wouldn't consider Wikipedia as being any more self-aware than my local library. It's the application of what knowledge you have that makes you capable of passing a Turing test, not the knowledge itself.

Yes and no. Humans have a certain knowage base. They also seem to have certain ways of link knowage together. Anything that is going to pass the turring test is likely to need these. Wikipedia can to a degree provide both. Of course wikipedia itself is not self aware but it is doubtful if a self aware computer could pass the turing test.

 

[mummymonkey]

Me - Hi
It - Hello there
Me - Where?
It - Where would you like?
Me - Like what?
It - You know what like!
Me - Do I?
It - " I don't know. Something."? Do you?
Me - Do I what?
It - Not that I know of.
Me - Saucy!
It - The members, or parts, of the starship USS Archon, which disappeared there one hundred years, until the USS Archon was a large, unfurnished room where Landru appears to his acolytes -- or, at least, a projection of him does..

 

[Angus McPresley]

Yes and no. Humans have a certain knowage base. They also seem to have certain ways of link knowage together. Anything that is going to pass the turring test is likely to need these.

FYI, the biggest effort to do something of that sort is a project called Cyc.

I was working towards a Masters in AI when I decided the whole effort was going nowhere. Climbing trees to try to reach the moon, as I've heard it said. Not that I think AI will never be attained; it's just a long, long way off. Also, the field is rampant with bullsh*t (the Loebner Prize awarded every year to the program that comes closest to passing the Turing test being one example).

Wikipedia can to a degree provide both. Of course wikipedia itself is not self aware but it is doubtful if a self aware computer could pass the turing test.

See now, that's the sort of statement you see in AI all the time. One big problem is that nobody seems to want to define their terms. There's probably no way to define "self awareness". I can write a C program that prints itself out, and tells you how bytes its code is. Is that self-awareness?

That said, if a program were to have a full, rich concept of self, by some definition, then it could probably have full, rich concepts of other things, and would be a long way towards being considered intelligent.

 

[geni]

That said, if a program were to have a full, rich concept of self, by some definition, then it could probably have full, rich concepts of other things, and would be a long way towards being considered intelligent.

Maybe but I still doubt it would pass the turing test. Aliens could be intelligent but we would not expect them to pass the turing test.