Artificial Intelligence

Will Artificial Intelligence
Ever Surpass Our Own?

By Mark Wherry...

When you think of artificial intelligence, do you immediately think of the phrase, "take me to your leader!"? And, if you do, is that all you think of?! Does it conjure up images of sci-fi movies that you are are ashamed to admit you have watched? You may be surprised to find out how much more there is to it.

A dictionary will tell you that artificial intelligence "is the ability of an artificial mechanism to exhibit intelligent behaviour", ie. a computer. But it is more than just that, artificial intelligence means giving a computer the ability to think for itself, to make choices of it's own accord, and to look at all the information and make a decision. The term artificial intelligence was coined in 1956 by John McCarthy of MIT. He defined it as "the science of making computers do things which if done by men would require intelligence." This was at a summer workshop and others who attended included Allen Newell, Herbert Simon, Marvin Minsky and Oliver Selfridge. Early work in artificial intelligence included attempts to simulate the neural networks of the brain with numerically modelled nerve cells called perceptrons. Success was limited due to the great complexity of the problem, but interest was revived in 1980, and has continued into the 1990's with new advancements in technology. In the late 1950's and into the early 1960's, Newell, Simon and J.C. Shaw offered a "logical theorist" computer program, and introduced symbolic processing. So, instead of building systems based on numbers, they attempted to build systems that manipulated symbols. Their approach was powerful and is fundamental to most work in artificial intelligence today. In this method knowledge is understood by rules, for example, "If X is a bird, then X can fly." If such artificial intelligence system is told that a robin is a bird, then it will tell you that a robin can fly.

However, even with these basic principles of reasoning, isn't there a danger of a computer that could think for itself, what if it went out of control? Such a system might start killing people, or decide not to do something we needed it to do, surely it is better to stick with our current ideas where we control the computers and not vica-versa.

But does this mean that we are afraid that we might create an entity that WE cannot control, is that not selfish on our part? Aren't we already heavily dependant on computers, imagine if there was a power cut, what would happen in offices, libraries, schools, banks or other businesses, the world would come to a stand still. Is this not already computers controlling us? We have already become too dependant on them to manage without them. But surely it is better to explore all of the possibilities, imagine if our artificial intelligence could then be recreated in a robot, then it could help us with many other tasks than just thinking. The theory of automata deals with the fundamental behavioural principles of automatic machines. It provides the theoretical foundations for automated processes in which sensors and detectors replace human sense organs, actuators powered by electric motors or hydraulic forces replace human muscles, and microprocessors replace the human brain. it could eventually be another human.

Do we really need another human and even if such a device is created will it just turn out to be another Frankenstein? It is true that we have become dependant on computers, but people always managed without them before. Also, there is a difference between dependency and being dictated to and how can we be sure that artificial intelligence in human form will be all bells and whistles what if there is just one problem that isn't ironed out? We have seen consequences of these thoughts portrayed in such films like "Terminator" where robots are trying to exterminate human life, isn't it worth exploring all of the possibilities?

It is not fair to always associate artificial intelligence with computers taking over the world, if it were possible to create such a device it could be programmed with information not to 'kill people' and if a fault occurred it could just quietly shut itself down. But perhaps this is looking too far into the future. In 1936, Alan Turing, an English mathematician, systemized automata theory by developing, on paper, a model for the digital computer. He proved that his "universal formula" could solve virtually any mathematical or logical problem that could be formulated in a logical manner. He also showed that certain types of mathematical problems were beyond even his method. In my view I would call this process linear artificial intelligence as the thinking has to start at one point, work through a series of processes and come to the end. The problem is that it can't jump around, working out different parts at a time, or thinking of other things at the same time. This would require what I would call non-linear artificial intelligence, in other words thinking that didn't have to go in a straight line or have a starting and finishing point. Thinking that could be free like a human brain. But, at present we have simple examples of artificial intelligence in everyday life, such as computer chess games; where the computer has to view the consequences of each possible move, likewise in problem solving; for example, a program that can solve anagrams, or in complex databases; examples of which are in the field of medical or mechanical diagnostic where the computer assists a person in deducing what is wrong with something. As we all know, such systems are invaluable to us, so why do we fear or condemn future development or achievements with this technology, and why do some people disbelieve that computers with a mind of their own is possible.

Alot of people when asked about such matters express the opinion that through further developments in this area more people will loose their jobs, and this is true. Isn't it better to have people employed doing jobs rather than a machine? I would think so. The problem is that if there were ten people doing a job, they could be replaced by a computer and one person to maintain it, and that person could be maintaining lots of computers, and nowadays he could be controlling it on the other side of the world using a telephone line on the internet.

One of the biggest problems with artificial intelligence is the amount of work, time, money and effort it takes to create a perfect artificially intelligent system and no one has yet managed it, because it is so complex. Alan Turing, a mathematician who wrote many papers in this field devised a vigorous test which any machine must go through if it is to be named as artificially intelligent. This test was named after him so it is known as the "Turing" test, and, to date nothing has passed. To give you an idea of how complex such a program would need to be, here is an example. When a robot tried to build a tower of bricks it had difficulty because it just dropped each brick from the same place because it didn't understand how to put each brick on top of each other. That anecdote is actually true and taken from the book, "The Dream Machine". It illustrates that every human concept and ability must be programmed in order for it to learn. And this is one of the main problems in current ways of thinking in artificial intelligence, how do you write a computer program that can learn?

Because no machine has passed the "Turing" test yet, doesn't mean that no machine will ever pass it, we just have to wait for the next generation of technology. Already new products and technologies are being developed we just have to see what happens. At present all computer chips use silicon, and because silicon is a metal and conducts electricity none of the paths in a chip can cross because there would be a short circuit. However, scientists are developing new chips which use light instead of silicon, and the paths of light may cross without causing any damage. This means chips will work faster, but not just faster, they will work at the speed of light! To give an example, with silicon if you wanted to travel from London, to Birmingham, you might have to go via America, however, with light you could travel straight to Birmingham.........Now, just think of that speed increase for a moment, it is incredible. And, of course, faster chips mean a computer can do more calculations in the same amount of time. I would agree that writing the program for artificial intelligence is difficult, but not impossible. All that is needed (theoretically) is a complex database program, and as the computer does things this database is updated and then at every decision the computer has to make it searches through its database to find a solution, much like a human will. A program to achieve this is really just a complex continuation of simple artificial intelligence programs we have already like problem solvers such as spell checkers and anagram solvers. In the instance of anagram solvers the computer has a dictionary of words which it has learnt and then unscrambles the word and then checks the word against the dictionary it has already learnt and reports to the user when a match is found.

How long should we wait for the technology to achieve this, and do we really need it? Infact today several scientist are beginning to question whether the usefulness of artificial intelligence is being over done. Is such a system, when it arrives going to be worth all the money that was spent developing it? What happens if we do get it, what next, how can it be kept under control? We may be told that it will be safe and that it could be programmed not to harm anyone but if this program is evolving and the computer is rewriting it's rules how can it be stopped from evolving out of the rules it was given by humans. If we are comparing intelligences humans have learnt to kill why can't an artificial intelligence learn to ?

How long we wait is irrelevant in the long term, how long do we wait for anything? How long is a piece of string? If we have an artificial intelligence then it would be able to do anything our intelligence would be able to. Give it tools in which to operate with ie. hands etc.... then it would be as capable as any human and be able to do any job. If we created something like an intelligence then we would be able to control it and it could be written with fail safes so if it could be programmed with things it couldn't learn. This would spoil the evolution a bit, but would enable us to keep control.

So what is your opinion? Does it really matter on what our intelligence is based, whether it be carbon or silicon? I don't think so. It is easy for anybody to say it is not possible but a hundred years ago if you'd of been told about the microwave oven, would you of believed it? With that in mind today we have technology people wouldn't of even dreamt about a hundred years ago, so it is difficult for us to know what will happen in the next a hundred years.

Will artificial intelligence ever surpass our own? I think that artificial intelligence will eventually equal our own and yes, artificial intelligence, when it gets to that level will be able to surpass our own because where as we are one body a computer can be a network of bodies. And, each network could be linked to other networks to provide us with the ultimate intelligence right into the year 3000!

Books to read...

Programs to run...

This article was written for the Falcon FacTT file for their Atari Phile magazine, but the author retains the copyright. All trademarks are acknowledged.


Aus: Atari Phile 04 / , Seite
Links
Copyright-Bestimmungen: siehe Über diese Seite