Help, the machines are taking over... (Artificial Intelligence)

Name: MHWake

Hello doyoo user, You have to be logged in to use these functions... Login or register Close window

Send message to member

Product:	Artificial Intelligence
Date:	13/08/01 (400 review reads)
Rating:

Advantages: Pushes the boundaries of what is possibel with computer technology, Can help make best use of human knowledge

Disadvantages: We need to understand a lot more about consciousness first

The creation of Artificial Intelligence has been a goal for computer scientists for at least two decades. When Deep Blue beat a human champion, the world's press presented the story as a huge triumph for Artificial Intelligence.

I would have to admit that the computer that can't beat me at Solitaire, let alone Chess, hasn't been invented yet. (I don't see that in all the papers as proof that AI has arrived. I wonder why?)

What it isn't
*************
A common sci-fi paranoid scenario - its best recent expression being the Matrix - is the world taken over by robots/computers. The idea is that super-intelligent machines could rule us rather than being our servants. (You could trace this back to the ancient golem myths, through Mary Shelley's Frankenstein to most of the 20th Century's sci-fi film output - including 2001, War Games, and so on.)

If this concept is a metaphor, then it's useful. It probably has a lot to tell us about the nature of our society. For instance, we seem to allow the unbridled development of technology to dispense with our labour - even though the world is bursting with humans who need work and we humans can't earn enough to consume technology without working. We have allowed our societies to be taken over by machines. We don't seem able to challenge the direction of industrialisation.

However, the creation of machinery still rests in human hands. It's still driven by human decisions - whether made by governments or by the shareholders and managers of huge corporations. The impact of the continual advancement of technology is enormous. We struggle to grapple with how it will change our societies. However, once we stop understanding that its impact depends on human choices, we really are its victims.

So, if our fears about technology reflect our society's increasing alienation, then these fears are surely suitable subjects for works of art.
In that context, books and movies about artificial intelligence are part of the process of human beings trying to understand their creations.

However, the concept of machines being able to develop artificial intelligence and using it to direct our lives doesn't work at all outside the realms of metaphor.

Firstly, there is the major problem of human consciousness. Computers are not self-aware. They may be able to outperform us by carrying out computations (the same sort of procedures for which people have used an abacus for thousands of years) in an inconceivably short time.

But- adding one number to another can not constitute consciousness, no matter how quickly or how many times it is done. A computer is doing nothing until its electronic pulses are interpreted by an interface with a human being - the operating system, for a start. An ant has more intelligence than the most powerful computer.

The main distinction between artificial intelligence and biological intelligence comes down to a con trick. The Turing test for artificial intelligence rests upon a human observer not being able to tell whether s/he is communicating with a machine or a human being. This is obviously a difficult challenge.

Nevertheless, it is quite easy to simulate a simple version of a Turing test machine. I used Macromedia Flash to simulate a real-world ICQ conversation - if that's not a contradiction in terms - by creating an array of conversational sentences and using a random generator to present one of these sentences in reply to the user's input. Most people who have tried it have been convinced there's a real person on the other side of the message line. (Of course, using an ICQ/ chat-room scenario is setting this up to succeed because actual ICQ conversations are usually full of non sequiturs and new conversational strands. A serious debate would be more difficult, but would only need more care devoted to the programming.)


My point is that this is just a false form of AI, based on the knowledge of human expectations. It can only reflect the intentions of the programmer. It's like a stage magician's mind-reading trick.

If machines are to develop truly artificial intelligence, it could only be through the incorporation of consciousness. When there are biologically based computers, perhaps using genetic material for processing and storage, then we will have to start to worry. (Not last, because we human beings don't have a very reliable track record when it comes to playing God.)

What it is
***********
The most advanced Artificial Intelligence at the moment is found in the creation of programs that can learn from their experience.

As I have argued above, machines cannot "learn", in any meaningful sense, as they lack consciousness. However, they can appear to learn, if they are programmed to use their experience as data input. The program's "experience," in this sense, is its record of what input it typically gains, what usually happens after particular patterns of input, what effect its output has on the user's inputs, and so on. That is, as processing power and computer memory expand, they allows computers to incorporate more and more data into their calculations.

However, a difference in degree is not the same as a difference in kind. That is, no amount of repeating simple calulations will make a computer achieve lift off to consciousness.

The development of the "learning" capacity - located in the improvement of the skills of programmers rather than in machines - is helping to make games and other programs much more responsive to users and helping to make gameplay more realistic.

There is a specialised use of AI that may indeed threaten humans. The type of program known as an "expert system" 4 or 5 years ago - you never seem to hear this any more - a
llows the codification of human knowledge. It can be used to produce an "expert" knwoledge base that may replace your systems engineer, doctor, plumber, teacher, estate agent, lawyer, and so on, in many of the tasks they do. Such a system could learn from its successes and its mistakes, just as a human does. One difference would be that such a knowledge base does not die, forget things, have sick days or moods. Linked to a robotic device, it could even carry out many of the practical actions for which we need living systems engineers, doctors, and so on.

The economic logic of the spread of such systems seems compelling at present. We were once led to believe (by Tomorrow's World and similar technophile tv programs) that the 21st century house would have robotic houseworking devices controlled by a central computer.

This always seemed a little pointless. Housework - especially of the sort that could easily be done by robots - is poorly valued in money terms. We would need some convincing to spend even twenty pounds on a device that let us switch on the cooker from the living room. Housework is poorly paid, even when done as a job. For many years to come, it would be cheaper (and produce more reliable results) to get a cleaner for a couple of hours a week than to buy and train a robot house cleaner.

However, unlike house cleaning, professional advice is expensive, and lends itself to codification in expert systems. The spread of AI into such jobs seems economically inevitable to me. I even like the irony that the worst paid and least respected jobs - like cleaning, cooking, caring for children and the infirm - are the ones that are safest from the advance of technology.

********************

Yes, computers are taking over the planet, but not because of some innate capacity of machines to become conscious. They are taking over because we humans have not developed the ability to control how we use our own cre
ations.

We could certainly make much wider use of AI systems to help us towards achievements that are currently out of reach, such as space travel, control of diseases. But, we really should never forget that computer programs and data are barely even transferable from one operating system to another. We cannot be sure that computers incorporating AI can reamin useful for even ten years. It is really foolish to allow our enthusiasm for technical advances to override our basic non-artificial intelligence.

Summary: