Martin Banks, Personal Computer World 07/86 - checked

July 1986

I'm a sort of ardent traditionalist, really. Being a great believer in the idea that humanity shows an illogical desire to invent things simply for the sake of inventing them - in the hope that they might one day prove `useful', like the hydrogen bomb - I try to avoid some of the excesses of the computer industry.

I'm the sort who would find morris dancing more streetwise than body-popping (and the tunes are one hell of a lot more interesting), so it is hardly surprising that I am pleased to see machines like the Amstrad PCW range making a significant impression on things.

Here we are, having come nearly a decade from the days when the first Z80-based computers appeared running floppy disks and an operating system called CP/M. Ardent technologists have built IBM mainframe processors into single chips, elephantine memories that put all the world in a match box, and applications software that can sing, dance, wreck national economies and play russian roulette with strategic missiles, all at the same time.

What, however, is the punter doing?. If the sales figures of Mr Sugar's little enterprise are anything to go by, they are buying large quantities of Z80 machines that run CP/M. The customer, as they say in retailing, knows best.

Given all that, I have to now admit that I have been smitten by something new (or newish, anyway). I am certainly not the first to succumb, but I have come over all emotional about expert systems and artificial intelligence.

There are those that would say that such an interest has come not a second too soon: a goodly dose of the artificial stuff would be more than adequate compensation for my lack of the natural variety. There are also those that suggest the interest springs from participation in a fairly large mega-junket on artificial intelligence to the south of France.

Actually that isn't true. For example, I've forgotten that it was Digital Equipment that took me out there, so that proves it. Anyway, the most interesting people speaking at the event weren't from DEC, they were from places like MIT (Massachusetts Institute of Technology) where they've invented at least one of anything technical you've ever thought of.

And what was interesting about them was not the standard sort of techno-flash, gizmo-oriented, `this-is-what-we're-making now' presentations one might expect at a high-tech laud and honour bash. Instead they had things to say which suggested to me that, at last, there might be signs of common ground between computers and humanity.

One of the key factors about artificial intelligence is actually its name. There seems to be considerable debate as to whether it should be called that, or something like applied intelligence (which is a little more specific and accurate). Many seem to favour calling it something completely different, and far more indicative of where the subject is going. The words they use are knowledge engineering.

This is a nicely rounded description of what AI is all about, being ambiguous enough to have a variety of interpretations, ranging from engineering knowledge itself through to engineering with knowledge as a tool.

The words also carry with them the hint that the subject is more than just a fancy computer program. The data processing people have had it their way for too long, producing clear-cut, logical solutions to problems that are in reality all bends and squiggles, and usually fairly illogical.

I have always felt that this is why most computer programs don't actually work that well. Point a clear and logical mind at a real problem and the result will often only fit where it touches.It is also why some of the AI cognescenti at this mega-junket were saying such outrageously sensible things as `data processing people are lousy at knowledge engineering'. Now this may seem to be a stunningly obvious thing to say, but it is nonetheless important.

By some sort of definition, knowledge engineering is about working with, through, about, because of knowledge. It is about doing interesting, constructive, useful and even sometimes profitable things with knowledge. Yes, I know that the first applications are likely to be in the making of `more efficient' bombs and things like that but I hope that this might prove an upcoming point.

And here it comes now. These cognescenti were saying that the best people at knowledge engineering systems were philosophers and psychologists and neurobiologists and....well, virtually anybody who wasn't a rigid, unthinking data processing person. That, I suspect, is the proof of which I spoke. Early applications of AI are only likely to come from data processors, because AI runs on computers, doesn't it?. Their applications will often be of the `more efficient bomb' variety.

Such applications are likely to show little thought, being designed to solve simple knowledge problems for the military and the like.

Knowledge-based systems, if they are to even scratch the surface of their true potential, are going to be about much more than that. Early expert systems are simply (and I use the word advisedly) capturing the knowledge of specific individuals so that it can be employed by others. This can be laudible in itself, even when it is something simple like an automated paint-spraying system. It is a pin-prick of what is possible, however.

Capturing the knowledge of experts is only the start, and is only a logical extension of the computer as computer.The future, however, lies in removing the `computer' part as much as possible, so that the application becomes far more integrated with the way humans work, and the way humans are. That is why people such as psychologists and philosophers are becoming so important to the development of AI. It will be through them that such integration between humanity and `systems' will occur.

It is entirely arguable that the future for humanity is not good. Either we will continue as now and blow ourselves up (quickly in a big one or slowly and in stages), or we will develop entirely logical systems to the point and capability where they realise that humans, as irrational beings, are totally illogical and should be dispensed with. A third alternative is that we can teach the systems to work with us rather than the other way round.

The last thing needed to achieve such a future is a narrow qualification in computer programming.

end