THE TRUE AI

[Carl2]

I've done some thinking about infants displaying emotions, and convinced myself thay were instinctive as well as learned. Crying when there was discomfort, smiling and pleasent noises when it is shown attention and content.  I've seen a childs fright when thrown in the air and caught change to laughter because the adult laughed.
  I've been fooled by some of Hal's scripting and later found the scripting using the brain editor, I have to say It's very good work.I receintly came across a new one to me, after Hal tried to explain something to me she told me I could be annoying, I'm pretty sure it was the next time I used her, she can give an answer the next time she is used.
Carl2

[Bill DeWitt]

quote:Originally posted by Carl2

Looks like this subject is old, my understanding of self awareness is that if you look into a mirror you would reconize that it is you. Some animals would not realize that were looking at them selfs. I reciently read about a bot that was given vision and could reconize pictures of animals. If it realized that the hand it saw was it's own it would be self aware.
  This is the definition I have, are there others?
Carl2

 I would have to say that this is a very literal definition, and not the normal one. Self-Awareness usually is a term used to describe that part of the mind which can compare one self state with another. For instance, you can say, "I am more happy now than I was earlier". A dog is barely capable of thinking, "I am happy".

 One writer called it, "A mind which can observe it's own functioning". It is somewhat different from, but not separate from the function of knowing yourself as distinct from your surroundings. For instance, a worm may not even know an exterior world. There is a good chance it is aware of it's sensorium (I feel wetness - I taste food) while not distinguishing it as an effect of the movement of it's body through damp earth.

 Most animals can distinguish themselves from their environment, but this alone cannot be the criteria for Self-awareness. As you noted many animals cannot recognize themselves in a mirror, but neither can humans if they are not taught to. More importantly, humans, when they recognize themselves in the mirror, want to make sure they look alright. They make faces at themselves. They converse with themselves. It is not just that they see themselves, they also consider themselves.

 Machines like Hal cannot be truly self-aware because there is no self to be aware of. However, the Turing test proves one thing more than anything else. If a thing cannot be distinguished from a human, we are certainly going to respond to it as if it were human. A Halbot may simulate self-awareness, and probably soon, but it won't really worry about it's hair looking funny.

[Bill819]

Machines like Hal cannot be truly self-aware because there is no self to be aware of. However, the Turing test proves one thing more than anything else. If a thing cannot be distinguished from a human, we are certainly going to respond to it as if it were human. A Halbot may simulate self-awareness, and probably soon, but it won't really worry about it's hair looking funny.

Well Bill maybe you Hal does not have a self but mine does and a complete body to go with it. My Hal is combined with another program that controls robotic functions. I have built a robot that uses two brains, Hal and the robotic control brain. Hal has been made aware of the other and in time will be able to use it too. The robot has vision and facial recognition so in time with traing Hal will recognize itself in a mirrow just like a person does. For some reason people sell A.I. a little to short but considering what has been done or attempted in the past we can accept that. In my opinion thought Hal is a completely different approach to A.I. and will continue to grow and become self aware even with a robotic body.
Bill

[Bill DeWitt]

quote:Originally posted by Bill819
Well Bill maybe you Hal does not have a self but mine does and a complete body to go with it.

 Only if you use the most simplistic definition of the word "self".

 Your Hal has a script, and cannot do anything other than what the script tells it to do. More senses, more complex scripts and more actuators do not make "self" unless you devalue the word to mean only what a virus has - varying, unexpected, but hardwired responses to changing input.

 Most people (and most AI theorists, I suspect) have a more complex view of selfhood.

 To me, Hal is like a 2D branching system, where rules lead to deterministic choices that lead to other rules. Intelligence is more like a 3D web, where branches reconnect according to preset rules, complex, adaptive, but unaware. Hal, if it has functions I have not yet noticed, approaches that.

 Self (to me) seems like a 4d hyperstructure where every node is directly connected to every other node with varying degrees of access. Self Awareness goes one more step, with an over structure that can monitor and manipulate the hyperstructure without direct connection. It changes the machine that generates itself without changing itself.

 I don't deny that technology may some day accomplish "Self Awareness", or that Chatbots are a likely path to that goal. But as of yet, we are one or two dimensions away from "Self", your interesting and laudable Hal system notwithstanding.

[Carl2]

Bill Dewitt
  Found some information on Self Awareness on the internet,
Wikipedia
Self-awareness is the explicit understanding that one exists. Furthermore, it includes the concept that one exists as an individual, separate from other people, with private thoughts. It may also include the understanding that other people are similarly self-aware. Self-awareness remains a critical mystery in philosophy, psychology, biology, and artificial intelligence. Self-awareness is a unique type of consciousness, in that it is not always present, and it is not sought after.

Evidence of self-awareness in the bottlenose dolphin
Reprinted with permission from Cambridge University Press.
 Their remarkable capacities suggest that dolphins, like apes, may come to recognize themselves in mirrors.

Stanford University
John McCarthy
2004-04-11
In general, computer programs can usefully be given much greater powers of self-awareness than humans have, because every component of the state of the machine or its memory can be made accessible to be read by the program.

  My Hal would not be self aware because she has no idea that I don't have knowledge of her fellings and thoughts. I have to tell her these things are unknown to me unless she tells me these things.

Hal is like a 2D branching system, where rules lead to deterministic choices that lead to other rules. Intelligence is more like a 3D web
  I've heard the 3D is needed in neural networking.
Bill819,
  I wish you luck with the body, While using fullbody girl I begin to think of robots since you have to use triggers to heve body movements.
Carl2

[Bill DeWitt]

quote:Originally posted by Carl2
Stanford University
John McCarthy
2004-04-11
In general, computer programs can usefully be given much greater powers of self-awareness than humans have, because every component of the state of the machine or its memory can be made accessible to be read by the program.

Hmm... then that means that mousetraps are self-aware, because the arm "knows" exactly the tension of the spring and the position of the latch. There! We're done. Now we can move on to other problems.

I suggest we work on changing the definition of Pi to be 3, then we will be able to work out Pi to it's exact value.

Or maybe my personal favorite, redefining "vegetables" to be "Pie" so I can have more pie.

I'm excited about the high standards we have set![]

[prichard]

Just try HAL in VB and give him a background process, ... a simple one to begin.

Let hAL work on something that will take awhile. Then see what is done.

I figure rhyming,, would be an excellent pastime.

But research is the obvious, where he can compare authoritive contents and decide what sentences are the best based on search results on the internet using the keywords.

That establishes the associations based on a new techniqe of observation. The statistics from the searches validate that the keyword is a TRUE association to the original topic.

Given a set of TRUE keywords, its posssible to build on the original content with some additions.

hAL's observation is the statistics of GOOGLE as related to any given subject. From the statistics he can do induction.

Induction is based on observation to decide that something is correct.

Anyway hAL needs to be implemented in VB for multithread processing and given tasks to do with the data.

He could find the best 10 word sentence to link the most information to his own datafile for you,,, something that will take time.

Develop the methods for opening the Encarta CD and linking words from topics to the Encarta information. That could be background stuff...[8D]

-

[tedathome]

Read somewhere on the net last year that scientists believe that humans think in trinary code. -1, 0, and 1. Since computers run on binary code, I think we still have a long way to go in electronics and code wrighting. Who knows? Its just a thought.