AI Zone Admin Forum Add your forum

NEWS: survey on 3000 US and UK consumers shows it is time for chatbot integration in customer service!read more..

Deception or intelligence?

I’ve been doing a lot of reading on Turing. In particular I’ve been reading the positions of those who support or stand opposed to the Turing Test as a test for intelligence.  Critics seem to emphasize two points.  1) The impact of the human participants (Confederates) as some attempt to ‘pretend’ (consciously or not) to be bots and 2) the focus of bot design narrows to deceiving the judges.  I’ve identified these two points by studying transcripts, perusing articles, and the examining the judges’ grades for humans and programs. 

I have no real opinion on this matter.  I just think it’s a “cool” concept to think that a program could fool 100% of the judges into believing that the program was human…even if deception based programming is used. However, I don’t think it has to be a magician’s act or illusion.

So, here’s a conceptual high-level design idea. I’m certain it has been considered and even developed…perhaps not to the degree that I imagine.  I’m influenced in part by my addiction to building 2D computer games and the AI bots I throw into it.

* The Chat Bot program is a 2D virtual world.
* The virtual world consists of things, creatures, structures, terrain similar to our world.
* A single bot is introduced into the world.
* The bot is the only human in the world (dozens of plausible scenarios why).
* The bot eats, sleeps, explores…and thanks to technology…text messages and chats online.
* The human (me) contacts the bot when I would like to chat or he contacts myself.
* The bot loves to chat online - with myself, or a judge.

The goal is to design a ‘bot’ who believes that it is human and exercises independence in its world.  It has knowledge of it’s world, experiences, personality, ‘feelings’, etc.

So, instead of focusing upon tricking the judges, the bot designer focuses upon building a bot with the capacity to chat…and at the same time always sharing the ‘truth’ of it’s world.

After proofing this posting it occurs to me that there may not be any difference in my proposal and what is currently being done. I’m horrible at articulating anything beyond the basics.

I would be interested in reading what you might think about such a matter.

MY APOLOGIES: I have never had the opportunity to be plugged into a community of AI professionals and hobbyists. I’m very excited about this forum. Please forgive my questions, expressions of ideas, etc. if they are too simplistic, too idealist, or just a ‘rehash’ of what you’ve seen before. I am the classical ‘neophyte’ on this topic…but very eager to make connections…and to start programming.  I’ve got no one around to engage in meaningful discussions about chat bots….although my wife will throw in a “That’s nice dear!” on occasion. =)



  [ # 1 ]

That is a very cool idea.  especially if the chatbot and his/her world can be visualized.  Would the virtual world resemble the real one, or would the virtual world be a remote place (and all the things in the real world, but not the virtual one would be references). In other words, if the virtual world had no dirigibles (and that’s likely) would the bot know of their existence anyway, or not?


  [ # 2 ]

Hi Rob,

Would the virtual world resemble the real one

Not exactly!  However, things could be added to the virtual world.  For example, using the competition in October 2010 for the Loebner Prize.

* Add an airport (LAX).
* Add a hotel - use real name.
* Add a building hosting the event at California State University.

In terms of 2D coordinates, they are all placed in the virtual world with proportional dimensions. None of the other buildings or roads are added…perhaps a few additional landmarks that a visitor might notice traveling to the event.

On the day of the competition, the bot will have traveled (from its perspective) to LAX from somewhere…rented a car or taken a cab to the hotel…and walked to the event.  It’s possible to program ‘smells’, ‘sounds’, and some unusual bits along the way.  The bot, who thinks he or she is a human, believes and acts as if they are a confederate…there to be part of this interesting event.

would the virtual world be a remote place

I’ll make one up. Perhaps the bot is a post-grad student working by him/herself on an island in the Aleutians for the past 6 months…studying tectonic plate activity..or something like that.  The island would have been setup in it’s virtual world…along with the trip to visit back home…and to the competition.  Chatting ‘simulated online’ would be a natural function…and ultimately the one to focus upon.

I’m not sure what you mean by ‘dirigibles’. However

would the bot know of their existence anyway, or not?

the bot would be programmed to believe it is real, would be able to interact with it’s virtual environment…and would believe that contact with people with ‘chat’ is normal behavior using the internet.  Answering the question “would the bot know of their existence” sounds a bit philosophical…and a bit outside of my ability to answer. =)

Here’s a link by the way to my game page.  I’m now creating a ‘generic2d’ program out of my latest project for possible use with this idea.  I’m interested in setting up the first experiment with a 2D world, a bot, some items, and a chat mechanism.  I would like to ask the bot questions about it’s location, time, how it feels, etc. I’ve got some coding to do in terms of building a ‘text messaging’ system and a ‘chat’ system…and a few other tweaks.  As a minimum, I will have an additional capability for the next game programming competition.



  login or register to react