My thoughts on A.I. (Artificial Intelligence)

Member Name: The Duke

Product:	Artificial Intelligence
Date:	11/10/01 (306 review reads)
Rating:

Advantages: Beneficial to society

Disadvantages: A long way off

What is Artificial Intelligence (A.I.)? Most people, while having their own idea of what it is, might have trouble actually describing this to other people. These same people might also find their idea of A.I. to be different from everyone else’s.

A definition I’ve pulled off the World Wide Web states that A.I. is “The sub field of computer science concerned with the concepts and methods of symbolic inference by computer and symbolic knowledge representation for use in making inferences. AI can be seen as an attempt to model aspects of human thought on computers. It is also sometimes defined as trying to solve by computer any problem that a human can solve faster”

Well, that’s a bit of a mouthful, isn’t it? But the latter part of the definition is what most people think of. We’ve all seen a science fiction movie that includes some sort of “artificially intelligent” being, and naturally, we’re all concerned that these things won’t rebel and use us poor souls as it’s slaves like they seem to do in all sci-fi movies.

In films such as Terminator and The Matrix, machines become intelligent and the human race suffers as a consequence. Is this just a good sci-fi plot scenario, or could this actually happen? Can we all expect to spend the rest of our lives mining, and producing silicon chips so that some grand A.I. entity can procreate and rule the planet?

My point of view is this: I wouldn’t worry about it. I don’t think we’ll see proper A.I. machines within the next thirty or forty years. Look at what we have now – some sort of Eliza type program, and perhaps some sort of recreational A.I. (e.g. “Creatures” on the PC). Eliza was a program that you typed questions to, and it would reply, and in turn ask questions back. Early editions of this were exceedingly weak, and you could spot the fact that it was only a simple program very e
arly on. Some people may argue that Deep Blue is a step forward in A.I. (for those of you who are unaware, Deep Blue was the computer that defeated Grandmaster Gary Kasparov at chess). Others simply believe that Deep Blue was a “number cruncher” with some nifty programming and not at all intelligent.

Years ago, Alan Turing, arguably one of the founders of computer science claimed that by the year 2000, computers would be able to pass the Turing test at a reasonably sophisticated level. I think he clarified this to be that the “interrogator” would not be able to identify a computer more than 70% of the time after a five minute “conversation”.

The Turing test is basically where someone (the interrogator) has contact with someone or something else (they do not know if it is human or a computer) and during a short conversation where they ask and get asked questions by this unknown other, they must make a decision on whether they are talking to a proper person or an artificially intelligent machine. Eliza, as I should have stated above, is a program which tries to beat the Turing test, but is quite old (the first versions are over 10years old) and was quite obviously not human!

Now, it’s 2001 (obviously!) and it would appear that Turing was way off the mark. What will have to happen before we can get a totally artificially intelligent machine? Well, first of all, it depends what you want the machine to do. Machines with simple tasks only have to be intelligent in one area, but if you wanted an artificially intelligent robot like in Terminator, Bicentennial Man, A.I. etc. then you’re going to have to be able to mimic everything the human brain can do (ignoring all the bits about building the actual body etc.)

There is research going on all the time, all around the world, into A.I. It should be noted that the majority of this research is into specific areas of A.I. and not immediately tryin
g to produce machines that are capable of thinking like we do.

There are many branches of A.I. Some are regarded as ‘problems’ and some are regarded as ‘techniques’. The research into A.I. problems and techniques being conducted around the world now would include stuff like:

Natural Language processing (NLP) – the processing, and understanding, of natural human language.

Automatic Programming – describing what a program should do and have the A.I. system write the program.

Machine learning – producing machines which learn from experience.

Visual Pattern Recognition – reproducing on a machine, the human sense of “sight”.

Speech recognition – the intelligent conversion of speech into text.

Neural Networks – the study, and production, of programs which function in a manner similar top how animal (specifically human) brains.

Now, as you can guess, it would take all of this research (plus more on other topics I have not mentioned) to be able to produce a complete human like intelligence. You have to consider if there actually is such a place for a human like intelligence though. As well as that – you have to try and consider what the memory requirements would be. A recent (1996) study by Dr. Chris Winter indicates that an entire lifetimes memories would need around 10 terabytes. To put this into perspective for you, your computer probably has around 128 megabytes of memory. Now, one gigabyte is 1024 megabytes, and one terabyte is 1024 gigabytes. So, you’re looking at computers which would have around eight thousand times more memory than your current PC.

But that’s not to say that if we combine one, two or even three techniques, we can’t come up with some good “real world” uses for a specific type of limited range A.I. machine.

What about:

Autonomous Vehicles? ̶
0;Intelligent” vehicles that drive themselves. Obvious benefits include – driving home from the pub safely! Having automatic drivers, programmed not to break the law might even reduce traffic fatalities.

InfoBots, KnowledgeBots or WebAgents? Online “bots” that are able to understand, and react to things that are “said” to them. These could obviously be used for all manner on things like monitoring of email traffic, web searches etc. If you combine this with advanced speech recognition / processing, then you could have better help desks for solving problems.

Portable Translators? Pocket-sized computers that can understand what you say, and then translate into a language of your choice might also be a handy contraption to have.

Program Debuggers? Programs that are specifically designed to iron out the problems caused by human programmers. The other side of this is an A.I. that can write programs quicker and more reliably than humans, especially for mission critical software like autopilots on planes.

A personal favourite of mine is one that analyses your telly watching preferences. Can you imagine being down in the pub, only to be told by text message that you’ve got 45 minutes to get home because the new series of Banzai! Is about to start, and your super A.I. telly realises you’re not in the house? It could ask you if you wanted it taped to watch later!

Some uses I think are simply pipe dreams. Many people have suggested that if we manage to come up with a full A. I. that was very like the human brain, and then scientific advances would happen very quickly because you could ask an A. I. to work out complex scientific problems and theories. It could then be assumed that seeing as the A.I. is not a real person i.e. no need for food, sleep etc. then it automatically could solve problems at least twice as fast as a human, but if you consider than processing power can be much fa
ster than a human’s, then theories could take weeks or months rather than years to formulate and prove.

These are just some of the uses I can see for A.I. off the top of my head. I’m sure if I sat down and gave it some thought, I could come up with more. I’ve tried to think of uses that apply to the real world, and some like totally autonomous planetary explorers where NASA etc. drops these vehicles on Mars and lets them do their thing, and report in when necessary are also useful, but to a more specialised community.
Now onto another important point which is associated with A.I. – emotions. Will A. I. entities ever have emotions? My opinion here is “no”. At least, not in any way that we’ll ever be able to understand it. My point above goes some way to explaining why I think machines will not emote. What’s the point in programming a machine to be able to feel emotion if, at the end of the day it’s only a set of eyes or ears and a limited intellect?

I suppose, if we ever get around to creating totally artificial beings modelled after ourselves, then there might be some argument for programming (emulating, in other words) emotions into them, but for now, I can’t ever see that happening. What possible reason would we have for teaching, or programming, machines to love or hate or even feel guilty? If we make machines self-aware, should we then give them a conscience – a sense of right and wrong, a set of moral values?

Is a set of programming commands actually an emotion though? As humans, there are some physical side effects to feeling an emotion – when we’re embarrassed, most of us flush. When we’re in love, our hearts “jump” when we see or hear the object of out affections, when we’re scared, our senses are heightened – without these physical components, can we really describe emotions?

Programming languages are based on l
ogical steps, and to solve a problem you simply go through a series of steps until a result is obtained. Trying to define emotions will probe to be an exceedingly hard task indeed, if it is ever decided that emotions for artificial beings are necessary.

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

So, there you have it - my thoughts on artificial intelligence. I’ve tried to keep it as simple as possible (at least in terms of technical language or jargon) and I’m sure I could have gone on much longer regarding the possible uses of A.I.

It’s something I don’t think we’ll have to worry about for a long time yet – computers are getting more powerful and programs are getting larger and more complex, but we still don’t give them complete control over our lives. Even my proposed autonomous car is still a long way off, if the automatic pilot on a plane is anything to go by – we still don’t give them complete control, and even when they are in charge, it’s only to do simple things.

If you imagine that an automatic pilot is a complex thing to program, and it’s capable of one task only, imagine the amount of programming work that would go into making a complete machine that can fly planes, drive cars, visually recognise thousands of different people and objects, aurally differentiate between many different sounds or voices as well as have the ability to learn, and form it’s own thoughts and ideas. Just like you and I can now.

Summary: