Artificial Intelligence

Bella wrote:Also, the major point on A.I. were it ever fully realized is that nobody would want it;

Slow that cart down please. I'd like to speak for myself. I'd really like to have my very own Cherry 2000. She could look like Alice Truman or Kelly LeBrock and I don't care what kind of Weird Science has to produce her. And I wouldn't give a shit if she wasn't intelligent at all.
http://youtu.be/kckEEQKXaCU

JohnRoth wrote:The basic fact is that we really can't define what "intelligence" is, let alone define what it would take to build a computer system that would do roughly the same thing.

I wouldn't attempt to build an AI to be intelligent. I would attempt to build an AI to simulate human abilities. Intelligence is like love. Metaphysical. You are not going to capture the metaphysical within a box. Just like love, there is no absolute intelligence. Unless, like myself, you believe in God. I don't think we're going to be able to design an algorithm with God-like abilities. (although as a computer programmer I often refer to the init process as the god process).

We cannot define intelligence, because just like love, it is infinite. We do not have the capacity to understand the infinite. Humanity thinks it's intelligent. Our uninvited UFO friends I'm certain beg to differ. If we are being visited by aliens, do you think they think we're intelligent? A species that kill each other over skin color? Religion? Who bang their heads on the wall to music? Why do you think they've never landed their spaceship? "I'm not going down there!! You're as crazy as they are!!!"

What it would take before any attempt to build an AI, would be to first overcome a major stumbling block in programming.

Gödel's incompleteness theorems and the Halting problem.


An aside:
It is interesting that one of the oldest computer languages in existence today was created for the sole purpose of AI research. It is just as interesting that it is undeniably the most powerful computer language in use today. No other language even comes close. There are things that can be done with Lisp that simply cannot be achieved by any other language. It has been and still remains my language of choice since I was a preteen. Simple. Powerful. Elegant. It hasn't changed much since 1958. It doesn't have to. Computers are still playing catch up. It is so powerful, that in the 70's, special computers were built to run it. Lisp Machines.

I have no doubt about the language that will be employed if such an undertaking of AI is achieved.

cralkhi wrote:Even if there's nothing necessarily "bad" about AI I think it would be a truly dangerous technology simply because if an AI thought differently from a human, we wouldn't understand it well enough to "raise" it and it would probably end up with psychological problems.

(And if it DIDN'T think differently from a human -- what's the point? We can already reproduce humans.

Similarly for human cloning -- I really don't see the practical application. cloning individual organs for transplants, sure. People, no point to it.)

Oh there´s plenty of points for cloning, they´re mostly just not very NICE points.

Try reading "Divine Blood" on FFN, where Kodachi Kuno makes a perfect showcase just how nasty some of those points can get.

Ch39 wrote: "And Kodachi Kuno," O'Neil said.

He clicked a button to show a dark haired woman walking determinedly through a plate glass window on the video and dying afterward.

"And Kodachi Kuno," he repeated, moving to another video. "And Kodachi Kuno. Ooo, gotta hurt! And, Kodachi Kuno..."

No, it´s not a realistic showcase, but it still flaunts some of the reasons why it could be a really bad thing.

Interesting discussion from a founder of the AI discipline.
http://youtu.be/Ozipf13jRr4
Dr. John McCarthy

Chthia, what you are talking about is called weak artificial intelligence. Weak artificial intelligence is definitely possible; we have the beginnings of it already. And there is clearly no problem with weak artificial intelligence. Basically you are saying you don't want strong artificial intelligence. Whether strong artificial intelligence is even possible is an open question. And you are not the only one who has reservations about the wisdom of actually developing strong AI. On the other hand, the search for strong AI has produced quite a lot of progress in weak AI.

SWM wrote:Chthia, what you are talking about is called weak artificial intelligence. Weak artificial intelligence is definitely possible; we have the beginnings of it already. And there is clearly no problem with weak artificial intelligence. Basically you are saying you don't want strong artificial intelligence. Whether strong artificial intelligence is even possible is an open question. And you are not the only one who has reservations about the wisdom of actually developing strong AI. On the other hand, the search for strong AI has produced quite a lot of progress in weak AI.

No SWM. I am talking about strong AI. Weak AI isn't worth talking about. Weak AI has been a reality for years! Chess is a popular example. Very capable chess algorithms exist and can easily beat the average player, and there are chess algorithms that can learn from their mistakes and adjust to a particular player's style. I'm talking about mainframe level chess AI. Even the commercial ones have been made formidable. (And yes, I can beat many commercial ones too) But that is weak AI. The computer isn't thinking. It's acting within a predefined rule base.


A computer learns by rote example, but it cannot learn by inference. Inference engines are limited because a computer cannot decide whether a statement is true beyond its programming. What that leads to is a computer that falls into an infinite loop when confronted with the unknown, or will prematurely exit. Code monkeys (programmers) jokingly refer to the halting problem as a construct that will yield 'frigid' (looping endlessly catching no joy) results or 'prematurely ejaculate' (throw an exception, halting before resolution.)

Even the most unaccomplished math student knows not to keep dividing 1/3 in hopes that it will terminate. A computer hasn't a clue. (Halting problem) It cannot determine that a process, an algorithm will not terminate. By relation it cannot determine whether new paradigms, rules for a rule based system, has merit. A computer will never be able to say, "Eureka, I've got it!" Because it cannot decide (halting problem) if what it has is what it's got. These simple decisions form the basis of thinking. Thinking is the basis of strong AI. Intelligence.

Many programmers rely on an IF/Then construct. Which is fine. But if the construct does not meet the If clause a computer knows not what to do. You can add else/if, else/if infinitely with a few abstract constructs. Lisp is perfect for that. However, a human can reason that this road will be fruitless. A computer...cannot.

Weak intelligence has always been realized. Chess. Theorem provers.

Gödels Incompleteness theorems and the Halting problem are unconquerable problems for strong AI. Truly thinking computers.

The Halting problem isn't considered to be an obstacle for weak AI type problems. Consider chess. Mathematically, every conceiveable move and variation possible on a chess board, and it's optimal response can be preprogrammed. No guess work involved, no unknown problems encountered, no decisions will need to be made outside the box (rule base) The problem with this at the moment, (which is brute force simulated intelligence) is the limitation of speed. Even humans equate speed with intelligence. "He's a bit slow, he rides the small school bus with the light on top."

Quantum computers promise unprecedented gains in weak AI and I undoubtedly can conceive of quantum computer computation making unprecedented inroads into strong AI.

I posited, as a joke in a Computer Science course, that computers could only learn by difference, and not by inference. Referring to the difference engine and inference engines respectively.

In summary, the Halting problem is an insurmountable obstacle for strong AI, (truly thinking machines) not weak AI, because weak AI has finite computational sets.

No SWM. I am talking about strong AI. Weak AI isn't worth talking about. Weak AI has been a reality for years! Chess is a popular example. Very capable chess algorithms exist and can easily beat the average player, and there are chess algorithms that can learn from their mistakes and adjust to a particular player's style. I'm talking about mainframe level chess AI. Even the commercial ones have been made formidable. (And yes, I can beat many commercial ones too) But that is weak AI. The computer isn't thinking. It's acting within a predefined rule base.

Aaaand, it´s cheating your pants off. Take away the huge databases those chess "AI"s use and they become quite stupid.
Enough that even i have a chance to win on a regular basis against the most extremely powerful one there is.

Cthia, perhaps I am misunderstanding you. You said that you would not want an AI that was truly intelligent, correct? That means you don't want strong AI, doesn't it? That's what I was trying to say in my previous post. I specifically stated that it was not clear that strong AI was even achievable, which matches what you say about Godel's incompleteness theorem and the Halting pro9blem. So I think I'm saying exactly the same thing you are. Did I misunderstand you?

Tenshinai wrote:

No SWM. I am talking about strong AI. Weak AI isn't worth talking about. Weak AI has been a reality for years! Chess is a popular example. Very capable chess algorithms exist and can easily beat the average player, and there are chess algorithms that can learn from their mistakes and adjust to a particular player's style. I'm talking about mainframe level chess AI. Even the commercial ones have been made formidable. (And yes, I can beat many commercial ones too) But that is weak AI. The computer isn't thinking. It's acting within a predefined rule base.

Aaaand, it´s cheating your pants off. Take away the huge databases those chess "AI"s use and they become quite stupid.
Enough that even i have a chance to win on a regular basis against the most extremely powerful one there is.

You have no idea just how true your statement is. Quite a while ago I had a very good chess playing computer, a semi-handheld, that among its features was the ability to disable its book opening database. I found that mode very refreshing. It played more like a human right from the start. Everyone liked it. But you could kick its ass quite often, though having to work for it.

Edit:
It was more fun because, since it was playing 'free hand off-the-cuff' there manifested itself many occasions where one minute you are winning and the next you hear 'mate in two.' Because the computer, is never distracted.

If quantum computers are ever realized, and evidence suggests so, they are going to kick the ever loving shit out of most any chess player. Because since time won't be a constraint, they can be set at the Master Level without having to wait. Can you imagine playing speed chess on level 10 of a strong chess playing program?

SWM wrote:Cthia, perhaps I am misunderstanding you. You said that you would not want an AI that was truly intelligent, correct? That means you don't want strong AI, doesn't it? That's what I was trying to say in my previous post. I specifically stated that it was not clear that strong AI was even achievable, which matches what you say about Godel's incompleteness theorem and the Halting pro9blem. So I think I'm saying exactly the same thing you are. Did I misunderstand you?

Darn. Sorry about that SWM. I forgot to comment on that. Yes you misunderstood me, but only because I probably didn't make it clear. I was joking about the Cherry 2000 that I hope comes to fruition in my lifetime. I was jokingly saying that I didn't care if she were smart. Like in the movie. The Cherry 2000 was a success in AI. But it (that particular model...the Cherries) was specifically made to be a dumb blonde. I was just being a dirty old pervert.

Perhaps I failed to make clear that I have reservations of mankind attempting to create an intelligence. Intelligence implies free will. Free will implies the freedom to become a homicidal maniac. Mankind cannot control itself, how can we hope to create something that will control itself? And what intelligent entity wants to be controlled? If mankind creates an Artificial Intelligence, we will have to use ourselves as a baseline. We will have to teach it everything we know. And what we know ain't shit. Therefore what it will know will be shit. And sooner or later, shit stinks.

True intelligence beget intelligent decisions. Beget intelligent actions. A suicidal, homicidal species (man) is not intelligent. A species that destroys his own planet is not intelligent. Creating a truly intelligent entity, to live and observe mankind's unintelligent nature invites rebellion born of self preservation and logic. If an intelligent machine has to live on Earth do you think it's going to continue watching man destroy its home?

But I think it's a moot point. Because of the Halting Problem.

An aside:
Being a computer professional, I've met many professionals who disbelieve in the concept of a God. Yet they believe in the possibility and even inevitability of true Artificial Intelligence? A phenomena that is intelligent, that thinks for itself is a life form. It's interesting that man has the nerve and audacity to think that at some point he can create life, but that an entity long ago couldn't have beat him to it.