Monday, 2 July 2012

Evolution, Technology and Intelligence ...

An interesting article came my way today, courtesy the Postulant. Published in the Atlantic, it is by Professor Dennett of Tufts University. Under the title A Perfect and Beautiful Machine: What Darwin's Theory of Evolution Reveals About Artificial Intelligence, it explores some interesting aspects of what has become a much used and sometimes abused term in modern 'training' circles. He points out that both Darwin's theory of evolution and Turing' creation of the 'computational engine' are underpinned by the idea that competence does not require intelligence.

Yes, you did read that correctly. You see, as Turing proved, it isn't necessary for a computer to understand mathematics for it to carry out extremely complex calculations. Darwin also pointed out that the evolution or adaptation of any given species did not require a complex understanding within the species to make the necessary adaptations to become 'competent' for the new environment it faced. But therein lies a small problem. As everyone who has ever used a 'computational engine' to perform a complex calculation can tell you, when it goes wrong, it requires someone who understands how it does what it does and why it does it to repair it. That is one reason species sometimes become extinct - they can't fix themselves when something changes too rapidly, and adaptation takes far too long. Example, the various species of flightless birds which have vanished from habitats because they could no longer outrun, hide or eat the new species that hunted them.

The reality is, that without intelligence, it is difficult to pass on knowledge or understanding of complex matters. Computers are actually extremely stupid. They don't think, they can only regurgitate what is in their memory bank or what is fed into it to allow the processing of whatever formula it has been told to use. I can recall being incredibly frustrated when using an early sprinkler design and assessment program, that it would never tell me where something was wrong, only that the outcome was not correct. Something a bit more like Sir Terry Pratchett's Discworld "Hex" with its neatly written message of ***Out of Cheese Error***Reboot From Start*** would have at least told me it needed feeding!

Turing's great breakthrough in computing was to see that the machine didn't need to understand the process. Essentially the computing part is a mindless process. So is evolution. It happens, thinking about it doesn't speed it up, nor does it alter what is already in process. This is where modern computers and the whole concept of creating "Artificial Intelligence" part company. An AI 'computer' needs to think, not to compute. That is the nub of the problem.

For a computer to compute at present it doesn't need to 'think' in fact it can't, but that doesn't make it 'incompetent.' That is where, I believe, much of the modern 'thinking' on training has gone spectacularly off the rails. If I want a human to 'compute' things they must understand the mathematics. To be 'competent' they must think. Some of my readers will recall being told when a new 'training' scheme was forced onto the UK Fire and Rescue Service, that 'competence' didn't require thinking, you only had to be able to 'do' whatever it was. I'm happy to say that we, the dinosaurs of training, have been proved right, competence does require thinking, it does require knowledge and above all else it does require being able to learn from mistakes and correct them.

Darwin postulated, according to some (though I can find nothing of the sort in his words reading his book and his correspondence), that in order for us to have evolved, all that was necessary was a series of accidental changes in our genes over several billion years. No, I'm NOT a subscriber to 'intelligent design,' but I do believe that the whole evolution of all the diverse species and forms of living creatures that make up the biosphere and are all interdependent, cannot have been entirely accidental. Who knows, perhaps some ancestor of Darwin's finches did decide a different shaped beak would be advantageous and set about stropping his on a convenient rock. Then, perhaps having proved his point, convinced others to do the same... Sort of puts a different spin on evolution and intelligence, doesn't it.

As Professor Dennett says, Darwin and Turing both had to 'invert reason' to arrive at their conclusions. There is still a lot we don't understand about either computers or evolution, but we're getting there. I rather think the ultimate answers may hold a few surprises for everyone.

No comments:

Post a Comment