Touring test, chat bots & AI

Member
Posts: 749
Joined: 2003.01
Post: #1
I was reading about the touring test ( http://en.wikipedia.org/wiki/Turing_test ) and how Touring himself predicted that in 2000 a 200 MB ram computer should be able to fool 3/10 people.

I played with the allegedly most advanced bots alice ( http://alice.pandorabots.com/ ) and jabberwacky ( http://www.jabberwacky.com/chat-george ) and they're nowhere even remotely close.

And these bots were made by 1 person only, seems to me there is a lack of interest in this branch by the big corporations-agencies.

Maybe they figured that making a "reasoning" natural language system is close to impossible and they've given up.

If i had to guess I'd say in our lifetimes we'll never have a bot that passes a touring test.

What's your guess?

©h€ck øut µy stuƒƒ åt ragdollsoft.com
New game in development Rubber Ninjas - Mac Games Downloads
Quote this message in a reply
⌘-R in Chief
Posts: 1,237
Joined: 2002.05
Post: #2
I'd guess Turing would be disappointed.
Quote this message in a reply
Sage
Posts: 1,066
Joined: 2004.07
Post: #3
Personally I think the reason nobody is working to the goal is because it really has no good point. So what if a computer can fool you into thinking it's a real person? I don't see much value in that and I'd guess that a lot of other people feel the same way.
Quote this message in a reply
Member
Posts: 749
Joined: 2003.01
Post: #4
I think the point of the test is that for a machine to be able to fool a man it must have some kind of natural language reasoning.

Though I agree natural language reasoning is maybe not the most useful form of AI (a machine that just does calculations fast might be more useful for humans) it's a kind of intelligence that is intrinsic to man (not say like playing chess) and in some sense defines us as "thinking beings".

Aside of the fascinating implications I'm sure there would be many concrete applications where it would be useful and profitable to have a machine able to "reason like a man", interface for information, interface for (simple) client interaction (say in shops, online stores etc...)

©h€ck øut µy stuƒƒ åt ragdollsoft.com
New game in development Rubber Ninjas - Mac Games Downloads
Quote this message in a reply
Moderator
Posts: 3,570
Joined: 2003.06
Post: #5
I think the Turing test is pretty neato (and AI in general, hard/soft/whatever). I've spent many hours over the years pondering it and playing with the chat bots. I even started my own project on it a few years back. One of these days there will be a major breakthrough in AI, I am convinced of at least that much.

My personal hunch is that it is going to take some kind of program that can self-write itself. I don't know what the proper terminology for that is. I'm not thinking specifically in terms of genetic algorithms or neural networks as we already know them, but perhaps something a little different.

The initial developer will simply create the "DNA" so to speak, and the virtual neural intelligence will "grow" itself by programming in its own logic (something similar to common genetic algorithms, maybe, but I suspect different). It will probably need to evolve and grow over time, like human beings do, under the influence of an environment flooding it with constant information that they somehow make sense of -- the internet seems like a good source of information. But real-world visual information and human interaction will be needed too, I'm sure.
Quote this message in a reply
Sage
Posts: 1,482
Joined: 2002.09
Post: #6

Scott Lembcke - Howling Moon Software
Author of Chipmunk Physics - A fast and simple rigid body physics library in C.
Quote this message in a reply
DoG
Moderator
Posts: 869
Joined: 2003.01
Post: #7
AnotherJake Wrote:[...]the internet seems like a good source of information. [...]

umm, LOL?

More seriously, the most amazing feat of the nervous system is not processing *all* that information, but filtering out all the nonsense. People incapable of doing so usually end up in rooms with bouncy walls.
Quote this message in a reply
Moderator
Posts: 3,570
Joined: 2003.06
Post: #8
By information, I mean stimulus. Of course 99.999999% will need to be ignored. The alternative would be to carry the computer around with you wherever you go, and hold its hand walking across the street and sending it to school and changing its diaper.
Quote this message in a reply
Member
Posts: 749
Joined: 2003.01
Post: #9
If you think about it we have already passed the technological singularity in hardware. Computers-machines which build better computers-machines which build better computers-machines already at an exponential rate without much need for human innovation. We are already at an exponential growth in hardware power.

Now this still hasn't happened with software though, almost all of the software is man-written (I think...).

©h€ck øut µy stuƒƒ åt ragdollsoft.com
New game in development Rubber Ninjas - Mac Games Downloads
Quote this message in a reply
Member
Posts: 245
Joined: 2005.11
Post: #10
I read of an experiment where they evolved electronic circuits in an FPGA (Field Programmable Gate Array). A computer was given a goal it had to acheive (i.e. for this set of inputs you should get these outputs) and started with a set of completely random configurations. It selected the designs which came closest and combined them into new circuits and tested those, then repeated until it produced something that worked. The final design made no sense under our current understanding of physics, and was dependant on a sub-circuit which was not connected to the inputs or outputs (the circuit didn't work if that section was altered).
If the same applied to software written by machines then I hope I never have to fix their bugs!
Quote this message in a reply
Moderator
Posts: 335
Joined: 2002.04
Post: #11
From what I've seen of current civilisation, there are a bunch of geeks like us trying to make computers as intelligent as the most advanced humans and the rest of society appears to be intent on making humans as stupid as the earliest computers.

So theoretically we don't have to work too hard to pass a Turing test, just wait for people to get stupid enough Smile
Quote this message in a reply
Moderator
Posts: 3,570
Joined: 2003.06
Post: #12
LOLLOL
Quote this message in a reply
Post Reply