Friday 2 March 2012

How many trucks will your helicopter pull?

From clipart by clopartof.com and clker.com
Of course you wouldn't ask such a question - would you?

Are you sure? Imagine an inventor 150 years ago who was hoping for funds to build a helicopter. The idea of being able to move from anywhere to anywhere sounds exciting so any potential financier would need the idea to be approves by his transport experts who you will find at the railway station. They would ask questions such as "We have 10,000 soldiers who want to travel from London Euston to Manchester Piccadilly - how much would it cost to send them by helicopter?" When the inventor replies that he would send them by train the conclusion will be that he has no confidence in his own invention. When asked "How do you ensure that your helicopter stops when the signals on the line are are red?" the inventor's repost that this is not a sensible question is taken as evidence that he is totally ignorant of how the great British transport system works. Needless to say the inventor would not get his invention funded.

The point is that the railway system is a very good method of transportation for high volume repeat journeys between known fixed points while the helicopter is for much smaller scale movements where the starting points and destination are not known in advance. The two are tackling a fundamentally different problem.

In trying to describe CODIL I have had a similar problem to the frustrated inventor of the Victorian helicopter. If you are reading this blog you are obviously very familiar with some aspects of the stored program computer. Many (if not most) of you will have been taught the basic principals of how a computer works - and quite a lot of you may be doing jobs where a commitment to the way they work is essential. 

But, in philosophical terms CODIL is about as much like the stored program computer as the helicopter is to a railway train.

Computers were originally designed for highly repetitive numerical tasks involving well designed pre-defined algorithms. The underlying mechanisms were never designed to be human-friendly - and to most people the computer in effectively a "magic black box." The more complex the task the more effort has to be put into pre-defining the algorithms and it would be interesting to know how many billions of man years (including time spent up blind alleys and on training) have gone into creating the computer-dependent world we now live it. While the benefits may be enormous, getting there has not been cheap.

CODIL starts from the opposite end of the spectrum. It effectively starts on any task from a position of ignorance - with nothing pre-defined. It does not formally break information down in to program and data, and there are no global rules. In effect it works by comparing information which it has just received with copies of information it has received in the past and uses this make simple deductions, nothing more. The fact this works may surprise many familiar with the limitations of the stored program computer, but many tests have shown a wide variety of situations where it demonstrably works. The whole is set up so that the human who is inputting the information can see into the "white box" so that he can understand how the information is being handled. 

An example analogous to the inventor of the helicopter comparison might help to emphasis the differences to the two approaches.

In the 1970s I had the opportunity to show Professor Donald Michie some of the Artificial Intelligence related tasks I had been experimenting with using CODIL. He was interested but then said "Very interesting - let me know when it can play chess."  But the chess playing task requires there to be a set of very precise pre-defined rules which are used in highly iterative calculations within the framework of a regular array of pre-defined dimensions. CODIL was designed for situations where there were no predefined rules and where mathematical constructs as sophisticated as a formal array of unchanging dimensions lay outside its basic capability. .

Of course Donald was very interested in Chess, so I am perhaps being a little unfair, but the example illustrates the problem I had. At the time the Artificial Intelligence movement in the UK believed that applying high powered mathematically inspired tools to the kinds of things academics consider very clever was the answer and their approach may well be what you need if you are looking for ARTIFICIAL Intelligence. The problem was that at the time I was trying to get support of CODIL and the idea that to understand human intelligence you needed to "think simple" seems to have been totally alien to academic establishment.

This news items in today's i newspaper is relevant to the problem of thinking simple. Mathematics is just one of the skills people are capable of doing. There are many other skills which vary across the human population - such as the ability to play (or compose) music, painting, writing books and plays, and the art of leading other people, displayed by politicians, and other. For all these skill areas I am sure you could say "At least half of all adults lack this skill"  In fact I suspect that the vast majority of urban dwellers lack sufficient of the skills of a hunter-gatherer to survive in in a rain forest!

But it is likely that 99% of each of the skills I have mentioned is culturally related. If we want to know how the brain works we may well be able to get some clues from looking in depth at particular human skills - but what we observe may often be the result of culture and the associated training in childhood.

What I am trying to say in The Evolution of Intelligence is that we need to try and find the highest common factor necessary for basic intelligence across the spectrum. If you take language as something very special in evolutionary terms - and unique to humans - it is easy to assume that this is language is the highest common factor.

I am taking a different approach and saying we should include animals (who do not have language) in our search for the highest common factor - and when we do this language drops out of the "essential common features" and just become another skill (admittedly very important) along with mathematics, playing a musical instrument and surviving in a rain forest. This immediately makes the problem to be solved very much simpler.

Once one has identified the highest common factor the key question is whether what you have identified is capable of supporting every human skill you could name. The question you need to ask is whether the CODIL model is a good first step towards identifying the "highest common factor" and whether the extensive research carried out many years ago demonstrates that it has sufficient inherent processing power to support a range of human culturally learnt skills.

No comments:

Post a Comment