Computers and Intelligence ... what is intelligence

[Sip]

Ok since we have this new forum, and we don't want to exclude the non-geeks ... here's a philosophical question about computers that doesn't need any special computer expertise to answer (but it is not an easy question) ...

 

So what is the definition of intelligence?

 

A few decades ago, when someone was really good in chess, he was considered "smart" and "intelligent". A child that showed special chess abilities, would be considered a "genius" ... But now, as many of you probably know, computers can beat the best human chess players ... does this mean computers are "smart" or have intelligence?

 

If yes, do you think computers will ever be more intelligent than their human creators? How can that be possible? How can we make something better than ourself? If we make something better than us, we would have to understand ourself completely ... in order to understand ourself, we would need to have more intelligence than we have ... so do you see the paradox I am running into or am I just way off ?

 

If not (if you think computers are not intelligent), why did that old definition of "smart" change for computers?

[Azat]

I started replying twice to this message yesterday and both times the power went off in our neighborhood and I lost everything.

 

I will try to reply once more today once I get my thoughts together because I think it is an excellent question.

[THOTH]

Yes - excellent question. I don't have an answer for you (concerning the definiton of intelligence)...but I do have some thoughts on your dileama.

 

We can create things that are stronger and faster then ourselves - why not more "intelligent"? We already have computer databases that contain more information then any one person could ever memorize, and processors that can linerally process faster then we do...so what is left? Some form of slef awarness and adaptive learning perhaps. (We already have some of the latter I think)...the former remains a mystery (and perhaps has already been breeched - but the computers are keeping quiet...how would we know? LOL). Of course many Science Finction books have adressed this subject. One I would recomend (except that it is out of print) is "Blood Music" by Greg Bear...where the protaganst creates self replicating/improving nano organizms (initially) designed to combat disease in our bloodstreams. Well they evolve rather rapidly - to say the least...

 

Another thought - I think that our future contains man-machine hybirds (and machine-creature hybirds) - already there is a great deal of research going on in this arena - both with microorganisms and organic components for machines. In our near future we will likelyhave available various implants to assist us - mentally, hoormonally, chemically or otherwise...and I imagine (once more accepted) the degree of machine augmentation will likely grow (will we be the Borg? eh?). And then there is the organic components to machines...a very interesting subject...all creating some very interesting moral/ethical and other issues...both related to and in addition to many of the standard genetic arguments....yes the times (and us) are a changing...one can wonder if (purely) organic life such as us is only a stepping stone to future machine or hybird beings? My, my the dilemmas of the (so very near) future...and some of us can't yet handle women driving cars! (seriously - these will be very major issues we will face as a species in the very near future - impacting us in a great number of ways...)

[Hyeflyer]

Now I am not that up on these wonderfull machines but it would seem quite possible for the computer to be moe inteligant.Arnt these macines the combination of many brillant program designers thus making them a combined inteligance?

[Azat]

I think this is an excellent question. I just wish I had the two other responses that I wrote last night as I spent too much time thinking about it and now I can't because of work.

 

First I should mention that I have a hard time defining what is AI? People have tried to define AI for the last 40+ years and I don't think there is a clear definition yet.

 

Deep Blue(the computer that beat Kasparov few years back) had some AI, or did it? Some people argue that for computers to have true AI they have to have feelings as well and make decisions based on feelings as well. My answer to that is AIBO the robotic dog from Sony(Personal note: AIBO would me a good Christmas gift for Azat) certainly had feelings and it learns as from its owner to a point where people can see dramatic differences between different Aibos after few weeks with different families. Does AIBO have AI?

 

My personal opinion is that many things have some sort of AI but true AI is at least 50 years away. If we think about it, our most advanced robots today have the intelligence of an average cockroach. Just think about the Mars Rover that JPL sent to Mars 2-3 years ago. It would take it 1 hour to analyze the space in front of it to move 20 feet. Now you tell me, can a cockroach not analyze its surroundings faster than that and move 20 feet faster than that. I know I am simplifying things too much, but robots have years to go to catch up to humans.

 

With that said, I also think that Robots will someday catch up with humans. Humans can "think" at about 500 trillion bits per second and if Moore's Law(computing power doubles every 18 months) holds true robots should be able to "think" as fast as we do within 50 years. Now there are physical limitations to the Moore's Law because the thinnest layer in today's Pentium class chip is about 20-25 atoms and once atoms become smaller than 7-9 atoms current starts to travel on the outside of the wire and thus we have shorts. But this is where Quantum computing comes in(basically bouncing off lights off of atoms). Today's most advanced "Quantum" computers are able to multiply 3*5 and get 15(that was an announcement couple of months ago). I know you are going to say that any 5 year old can do that, but this is a true leap in computing. Take that 50 years form now and we can have computers that can think faster than a human.

 

The question of what happens to humans in 100 years is also interesting to think about. Do computers run the world? Or will we be smart to put some mechanism in computers that shut them when they try to do "evil"? I don't have the answer.

 

I do not follow AI much anymore. At one time I was very interested in it. Even worked on software "CASE" tool that would try to solve hardware issues in a big network.

 

Couple of sites to visit if any of you are interested are:

AI MIT Lab: http://www.ai.mit.edu/

ALICE chat bot: http://www.alicebot.org/ - a very basic chat bot that uses some AI

University of Zurich AI Lab: http://www.ifi.unizh.ch/ailab/

Michio Kaku site http://www.mkaku.org - He is probably the most famous physicist to deal with AI amongst other things. I really love the work that he does and he is an excellent speaker too.

 

Couple of books to read:

Visions : How Science Will Revolutionize the 21st Century by Michio Kaku (the is a great book)

Flesh and Machines: How Robots Will Change Us by Rodney A. Brooks - Director of MIT AI Lab

The Artificial Life Route to Artificial Intelligence by Rodney A. Brooks - Director of MIT AI Lab

 

---------

 

Sorry for the confusing message. I am writing it at work in-between projects. Hey at least I am getting paid for this

[Sip]

Very interesting thoughts guys. I agree HighFlyer that the key will be abstraction, i.e. each individual human will be an expert on a specific thing and many humans combined will from an amazing intelligence that together can create something that is much more amazing than a single human alone

 

Now to computers vs. Brain ... maybe I have a different perspective than the average person since I can see that today's computers are not really that intelligent. They have to be told pretty much everything ... they also have to be told every step on how to process things.

 

However, I was recently exposed to some research results on how the human mind works. More specifically, on how the brain detects objects that it sees. It is completely fascinating ... we still know fairly little but it's just amazing how complicated the process is and if you think about it how well humans can recognize different objects that may even look very similar.

 

A single neuron in the human brain is a very simple thing. But when you have billions and billions of them interconnected together, you get some very amazing abilities with extremely complex behavior!

 

Currently we are reaching the point of having billions of transistors on a single chip. However, their connections are almost always very well defined and do not change during the lifetime of a chip. So in this way, computers can do fairly regular tasks amazingly well. However, they will never (or at least it will take a VERY long time) till they can become as sophisticated as the human brain.

 

Our brain is NOT very fast (I think the neurons fire on the order of a few micro seconds or something like that). It will never be able to compete with even a 1 GHz chip that can do at least 1 BILLION operation per second (1 per nano-second). But the brain can do many many things at the same time and most important of all, it is always adapting and learning. The connections keep changing and so do the patterns to which it responds.

 

I think the Turing test is still a very good one. It was proposed by Alan Turing in the 50s. It is a well known subject among the AI (Artificial Intelligence) crowd. The test basically says, let a human interact with the "AI". If after a short period of time, the human is NOT convinced that it is dealing with another human, then it is NOT intelligent!!!!