Some Notes on "Is true AI possible?"

Everyone on this board seems to take for granted that true AI is not only possible, but a given with enough time. I would like to argue the counter point that true AI may be all but impossible.

Lets look at some basic facts first: Human beings are able to do things that can bring any computer to it's knees. Not only that, but we can do MUCH better and MUCH faster. For example, image recognition. No computer on earth can light a candle to the ability of a two year old child to process and recogzine images. Not only that, humans brains are amazing resilient to damage, change, and removal of individual nodes (ie, cells). For most computer programs, flipping any one bit results in complete failure. Beyond that, human brains are remarkably good at handling novel data without failure or damage to the system. Most if not all AI computer programs, no matter how good they are, crumble if you place them in a new environment.

More importantly, humans seem to have a odd sort of "knank" for seeing paths of action that will ultimently fail. Take chess as a simple example. Grand Champions have an amazing ability to look at a chess board and determine that individual move will spell defeat 20 moves ahead without going through every move in between. For a computer to do the same thing, it must search the entire move space to reach that same conclusion (which in chess is a mind boggling number of states).

Despite the amount of research into AI, no program has even come close to what we consider true thought. It is unlikly that one ever will in our life times, even taking into account Moores law.....

I think that with the right amount of time, ANYTHING IS POSSIBLE!!

An infinite amount of time will not help in all situations. I could give you an infinite amount of time to create a faster than light space ship, but that is just not possible given the theory of relativity (Note: I'm talking about a spaceship here. Not a single particle).

with this government, i woundn't be surprised that they did it already.

Originally posted by wilsone8
An infinite amount of time will not help in all situations. I could give you an infinite amount of time to create a faster than light space ship, but that is just not possible given the theory of relativity (Note: I'm talking about a spaceship here. Not a single particle).


But your theory is only true as long as the theory of relativity is true, and what's to say that that theory cannot be proven wrong is some period of time.

Quote originally by wilsone8
An infinite amount of time will not help in all situations. I could give you an infinite amount of time to create a faster than light space ship, but that is just not possible given the theory of relativity (Note: I'm talking about a spaceship here. Not a single particle).

We only have proof from haydrons for the theory of relativity, the fact that as they fall faster and faster from the lithosphere they gain in mass. There has been some research to disprove his threory but a theory is only a theory and they can be disproved at any point in time.

I just recently read the new novel, TAKING THE RED PILL, and I must say that it was worth the reading. Being someone who doesn't read that often, I was intrigued by the scientific, religious and philosophical aspects that were presented in this book. This collection of essays by various people working in various fields, really made me think and ponder about the integration of AI in our society. After reading the essay WHY THE FUTURE DOESN'T NEED US, I was somewhat disturbed by the facts that were presented. Taking that the author is an inventor and a scientist, the things he talked about seemed pretty factual. In his essay, he says that at the rate were going at, and with the help of genetics and nanotechnology, well be able to create machines that think like we do, and evolve, in about 30 to 40 years. He says that because we have such a great dependance on the machines, we will be forced to create newer, smarter mahcines that will eventually control certain industries. After reading this essay, to tell you the truth, I was slightly scared at the thought of our civilization being controlled by a slew of machines who have virtually all the power, and who hold control in their grasps. And the outcomes are terrifying. One suggest that all fields in which humans operate today will be engulfed with machines, creating millions of people with no jobs. Another is that, because we will have become so dependant on the new machines, we will blindly miss that the machines are taking over, and that they don't need us humans to survive. Without even knowing it, we would be getting extinct by our own creations. But in the first place, we are still the ones with the power and the control. We don't need to create such machines, but our futur might demand us to. We might also be forced to create these things to survive. The chances of the human race continuing to have a hold on power and control is very small. But it makes you think doesn't it? How we have the capability to indirectly kill our own species. By thinking we are doing good, were actually creating monsters that will someday find the way to enslave us and destroy us. Good night!

True AI consist's of 4 functions. Learn, Act, React, Relearn.

I know for a fact that I or someone else that is alive right now will write the proper code for these functions to grow into pure AI.

BTW, if any chess player could see 20 moves in advance it would do them no good. It turns into variable hell. Appropriate foresight is a solid 6 exchanges with additional moves to be able to CYA if necessary.

True AI is, I believe, a possibility and a probable outcome of our technological evolution. Already we have machines that can beat us at chess, can beat us at Halo and can generally do things we can't. Also, we already have machines which carry out menial manufacturing tasks. What is to say that we won't develop true AI one day, capable of thought and learning.
And I don't think it is something to be afraid of. So what if our menial jobs are taken away? We could create a society based on the arts and sciences, not on wasting our lives farming. I suppose though, that that would be unthinkable, bceause it would be a form of Communism to give everyone free food. and so on.

Originally posted by Osyris
True AI consist's of 4 functions. Learn, Act, React, Relearn.

I know for a fact that I or someone else that is alive right now will write the proper code for these functions to grow into pure AI. This is purest balderdash. [Using a polite word so I don't get censored.] You most certainly do not know this for a fact. You might believe it, but that doesn't make it a fact. You have resorted to the least reasonable of argument techniques, namely faith. Or "religion," as some might call it.

Points to ponder...
How will you represent knowledge?
How will you distinguish correct knowledge from incorrect?
What programming language will you use to implement this?
What design paradigm will you use? (I.e., how will you construct the analysis model and, from that, derive a design model?)

damn dude you are an ego brain. Read what I wrote again. Think about for a couple weeks, then run with it,

YOU CANT SAY FOR A 'FACT' THAT THE MOON ISNT MADE OUT OF CHEESE... BUT ITS A FAIRLY SAFE ASSUMPTION THAT ITS MADE OUT OF ROCK.


ONLY A HOLOCAUST OR A BAN ON TECHNOLOGICAL RESEARCH CAN PREVENT THE IMINENT CREATION OF AI.

RAARRRRRRRRRRRRRRRRRRRRRRRRRR



^^ should be appended to wasted's post

Originally posted by Osyris:

if any chess player could see 20 moves in advance it would do them no good
I play chess myself to quite a high level and if someone were to claim to see 20 moves ahead then that means that they can view most recommended moves in the knowledge tree to a depth of 20.... considering the number of possible first moves that can be made we can roughly cancel most of them and in some situations i.e. such as endings where their are relatively few moves that can be made it is actually possible to analyse up to a 20 move depth. If we are talking from the beginning of the game however then this is simple BS(excuse my french!) because no chess player in their right mind would ever try to analyse their first moves they learn a repertoire instead which is exactly what chess computers do also...

When you consider a depth search you MUST consider all possible options otherwise it is not a search of depth 20 but a search depth of 19...

and you are correct. However, that depth is simply non-existent in the chess world outside the use of computational devices. A solid 9 moves from current move is sufficent in both human and computer worlds as of now. Would it be crazy to think that in the next 30 years humans will be able to sufficiently analyze that depth in a timely manner? I don't think it would be crazy at all. 20 moves is good stuff if you have the time and the need for all that analyzation. Must be one hell of an opponent. Josh Waitzkin saw only 8............

True AI is certainly possible even though at the current state of technology you would need the power of several supercomputers
and of course the right programming which is actually exist now in very premature forms of video game chracters but is to simple to evolve beyond that like in red faction 2 where the enemy AI gradually becomes more difficult as it "learns" by gaining speed accuracy ect. but can do no more because it can only learn from which the information it is given so to say true AI is impossible is crap because it is certainley possible just that no one has really tried to do it aside from video and computer games or the goverment has and like everything else they think we arent ready for they lock away in some hangar at area 51.

They've also used AI in Expert Systems, chess programs, natural language processing and other such mundane tasks. There have been plenty of different uses of AI other than with computer games. For a long time there wasn't much interest in AI until the last few decades where much of the AI research has been involved in the chess program Deep Blue. I guess you can't rule out the possibility of AI. Just consider who would have thought in the 10th century that we would have cars, helicopters, supercomputers, discovered the human genome and even been able to send men to the moon! Never say 'never'!

I would firstly like to commend your notice of an interesting issue, but would like to respond in kind.

Let me first note the fact that the power of our computer doubles about every 18 months - give or take a month or two - Moore's law, for those who know what I am talking about. Further more, let me note that, since 1950 when the most primitive versions of what we now comfortably call computers appeared, their power and ability has increased by a factor of about ten billion - such a dramatic rise in technology is virtually non-existent in history.

Granted, computers as we now build them are very finite; someone earlier on mentioned quantum computers - there is nothing negtating the notion of AI in our current physical limitations as there are many ideas as to how to further increase computer power. The brute fact is that despite the notion that computers and human brains function differently, the alarming rate of growth in the power of computers could lead to a type of artificial intelligence; in other words, if a computer can reason, think, etc. does it matter if the manner in which it functions is different from us? We already know that the degree to which we can increase power is beyond our own foresight...I see no contradiction there. In fact, there are many reasons to suppose that, once self aware, they may function better than us.

People participating in this thread seem to believe that, simply by virtue of bigger-faster-better computer hardware, AI will somehow "emerge."

This is not the case.

Hardware is relatively easy. Software is very, very hard.

Before indulging your fantasies about AI, consider joining a professional organization such as AAAI (http://www.aaai.org/) or SIGART (http://www.acm.org/sigart/). (If you've never heard of these organizations, you are not even qualified to use the term "AI" in public.)

Study real AI for a few years and then come back here and tell me what you've learned.

You may have noticed me using phrases like "may lead to" thus insinuating that it is a possibility, not any sort of material implication. Therefore, an individual expressing limitations purely in the abilities of computers' power and my response has nothing to do with software, nor was I implying such. My response was to the notion that it was an impossibility, not to the notion that it would magically emerge.

It doesn't take a rocket scientist, or an AI practitioner to note the difficulties in developing software and I wasn't saying that it would be easy. But imagine what the world's most brilliant AI reasearchers could produce with the machines of the future, not to mention thier own progress in this research. There are many in such fields that do believe that not only AI, but sillicone based artificial life forms are very much a possibility. There are many, many articles that go beyond this narrow scope.

I am not familiar with such organizations, but will now look into them. (note - organizations based on the research of artificial intelligence seems to insinuate that THEY believe it to be a possibility) I am, however, more concerned with the theoretical rhealm of possibility - after all, isn't that what the Matrix was about?

NottheOne : Firstly what's your problem? People are allowed to voice their opinions.

I've studied for 6th months as part of my degree modules (only a first year though!) the fundamentals of AI and have discussed and heard the opinions of some of my lecturers thoughts on the future of AI. And yes there are some computer scientists that believe AI is not possible and some that believe that it is.

I'm intending to take further modules in it as the year's progress so perhaps you could say it's impossible but to be honest I don't think even and arrogany &$^% like you or even some of the leading minds in the world know what is round the corner... People who study quantum computers themselves even say 'If you understand quantum computers then you do not yet understand it fully!'

So pipe down with your arrogant talk NottheOne noone wants to hear it. Just because people here aren't as qualified doesn't mean they have no right to voice an opinion that's just egotism!!

I'll just say this....


AI is possible, and no you do not need a super computer. You need a software base. It is a simple base as well.

You don't needs tons of code. Its the 4 simple functions I described above. With 2 missing details that I am not going to get into for obvious reasons.... credit and money.

Basic programs already use static AI, and most of the people in this world doesn't even notice it.

AI will come out eventually, regardless of technological achievements between now and the time its created. The person who creates it will end up being a programmer who stumbles onto something oh so simple yet perpetually overlooked by current researchers of AI. And quite possibly made on an old beat up 386. :D

You probaly do not need a supercomputer for AI but for the "layman" that of a supercomputer as these computer geuines that have this unlimited coputing power yes you would need one or think in 5-10 years a houshold computer will have the power of one our supercomputers of today and then what will AI be like when quantum computers go past the 1+1=2 stage which they are currently at and go into mass production AI will be almost "human" like nevertheless if things get out of control it could easily turn into another Matrix situation. :( :D

ahhhh . the great dilemma of AI. They'll never be human. They will be perfect....... and thats not human. Perfection is a flaw as intelligently put by the architect.

I think we are closer to developing software with capabilities to think independently- oops! It's already happened!

For Lord of the Rings, software was developed so the CGI 'warriors' would move independently of one another, which is a movie making marvel. Almost everything is programmed, so the warriors react to what one another does- a complex program, but still just reacting to what they are programmed to do. However:
"In an early simulation, Jackson and Regelous watched as several thousand characters fought like hell while, in the background, a small contingent of combatants seemed to think better of it and run away. They weren't programmed to do this. It just happened. "It was spooky," Jackson said in an interview last year. "

One step closer....

(read the article at popular science:
http://www.popsci.com/popsci/science/article/0,12543,390918-1,00.html

ahhhh . the great dilemma of AI. They'll never be human. They will be perfect....... and thats not human. Perfection is a flaw as intelligently put by the architect.

Perfection is in the eye of the beholder.

perfection to one may be a complete abomonation to another.

The difference between today's computers and an AI is that today's architectures DO NOT know how to process an error. Even the simpliest bit if at the wrong moment will cause a fault. Human brain WORKS on supposed and finite information. you can't remember every particle of dust in your home, you can tell it's dusty and needs cleaning, though.

Simplified and globalized patterns in a world filled with tons of information is the start. You can tell a glass from it's shape , transparency, you can tell a car from it's size, shape, airshields, lights, and mostly, wheels! This is where intelligence resides. If it can work this way, it must be it!

I won't claim to have a vast wealth of knowledge about AI or computer programming, so we can get over that before I even begin. But if you know anything about humans or any other biological creature, you know that our bodies are compose of smaller units like our organs which are in turn composed of different types of cells and so on a so forth. It breaks down to a finite number of simple chemicals, particles, and cells, all which obey a finite number of rules. Everything works according to the rules which govern it- we just may not understand why, or understand what rules there are yet or how they interact, which is I think the only reason why we do not understand everything about the human body yet. But there is really only very few degrees of separation between a human body and a computer one. That might sound stupid, because how could you compare humans and computers. But when it comes down to it, its all particles obeying the governing laws of physics and chemistry. "How do you define REAL?"

You define real as what your senses percieve but then again your senses might also percieve that a mirage is real but it most certainley isnt or does by some power beyond our understanding simply dissapear once we get close enough to it? we may never know.

u see only what u expect to see. nothing more. nothing less.

Originally posted by Osyris
u see only what u expect to see. nothing more. nothing less. Maybe that's all "u" see. But "I" see the real world. Please don't try to reduce all of reality to stupid platitudes.

Romulus I agree, and Osyris has a point whether it sounds ridiculous or not- we have no way of knowing whether we ourselves are in a Matrix. There isn't even any such thing as objective reality or objective sensory input. In recent years studies have shown supposed "eyewitness" accounts of events to be very unrealiable, and totally based interpretations and expectations. You really have no choice but to trust what you experience, yet at the same time realizing that you have no way of really verifying that experience to be "true".