AI - part 2 - Intelligence
In his 1981 / 1996 (reprint) book “The Mismeasure of Man”, Stephen Jay Gould argued that there is no such thing as Intelligence.
There are two main thrusts to his view.
The first is something called “reification”, which is our tendency to turn abstract concepts into discrete, and in this case measurable, entities.
I used to think that I was cleverer than my partner because of a number of reasons: I did more cerebral work than she did, I was in MENSA, I read more intellectual books, I had a degree in mathematics rather than her wishy-washy languages, I was better at Sudoku, and, well, probably also because I was a man (world, please forgive me). Then one day I got a Nintendo hand-held console with a little game on it which was all about sliding coloured shapes around a screen in order to make sets, and she wiped the floor with me. I couldn’t understand this, I was convinced I would win hands-down as this was clearly a mentally-based game, but, no, she absolutely slaughtered me on it time and time again.
What I was guilty of was thinking that I was more intelligent than my partner, rather than simply better at some skills that we associate with intelligence and worse at others.
This is reification at work – the idea that somewhere inside our brains there is a “thing” which we can call “intelligence”, and which governs every aspect of activity which we think of as being “intelligent”.
There is absolutely no reason at all to suggest that such a thing exists. Our physical and mental skills improve with practice, rather than because they are driven by something hard-wired within us. We fall into the reification trap because skills tend to group together into similarity-sets, which leads us to think that each set must be driven by some sort of inbuilt mental quality. This is not true. The reason you tend to find that someone who is good at maths will probably also be good at Sudoku is because if we enjoy one activity, and become good at it, or perhaps because we are good at it, then we’re likely to enjoy something similar, rather than because we are born with an inbuilt ability like “numeracy”.
Measuring intelligence is an odd thing to do. It’s a bit like trying to measure a person’s overall strength by taking a weighted average of the strength of every muscle in their body. Someone who spends all their time in the gym will doubtlessly score well but most of us will have a potpourri of strong and weak muscles which will result in an average which doesn’t really have any meaning at all. IQ is like that too; unless your score is ridiculously high (and I mean well above MENSA levels), it doesn’t say all that much about you beyond “you’re good at some things and bad at others”.
These days, however, people are strongly motivated into wanting to believe that such a thing as Intelligence exists. They also want to believe that it can be measured and, I’m afraid, that theirs is higher than yours. This is the second and, I believe, more important point that Stephen Jay Gould makes in his book, and which is largely responsible for the controversy surrounding Intelligence and their tests.
Over the last hundred years or so, Intelligence has taken over from Strength as the principal justification for wealth and power. Even if an “intelligent” person has done nothing to demonstrate the successful application of their “intelligence” anywhere, the perception of their being “intelligent” is enough to allow them leap up the ladder, whether that ladder be corporate, political or social. With so much at stake, then, is it surprising that so much effort has gone into trying to control how intelligence might be defined and, more importantly, how it might be measured?
Until I read “The Mismeasure of Man” I believed in the sanctity of the scientific process. After reading it I realised that scientists are just as biased as the rest of us, and subject to the same emotional, and social, pressures. In a world where intelligence is synonymous with superiority, can we really trust anyone to behave in a disinterested manner when trying to identify or measure what intelligence might be? Can you really imagine such a person emerging from their studies with the conclusion: “Actually, it turns out I’m pretty thick”?
In the previous part of this blog I described how everything that we think, remember and communicate is based on a world view, and the difficulties of trying to reproduce that artificially. I now add to that the concept of bias, introduced by means of this little sojourn into our attempts at trying to pin down what intelligence actually is. Our world views are only partially shared, the other part is very much biased in our favour. Part of the genius of our ability to communicate, however, is our ability to be able to account for each other’s biases, as was demonstrated by Mandy Rice-Davies in court with her now famous statement: “well, he would, wouldn’t he”. I look forward to the day that an AI neural network can make sense out of that one!