Saturday, February 22, 2014

Outwitted by an Idiot

So I am sitting in the Wetherspoons, and in front of me are a couple of fruit machines. The sound is turned off, since Wetherspoons values peace and quiet.

The fruit machines, far from being annoying, are quietly making restful patterns with their flashing lights in a relaxing sort of way, and it occurs to me that the word for what they are doing is 'screensaver', which is a type of program that damages one's screen.

Back in the day I was a programmer of very small computers (yclept microcontroller), and I begin to wonder idly about how the machine is laid out inside, wires to buttons and wires to lights and maybe an 8051 and a few bytes of RAM, and Bob would pretty much be your uncle.

And I think that given a few hours with the design documents I could understand completely what is going on in a fruit machine. It is probably considerably less complex than a small 1980s motorcycle.

In fact, unlike a motorcycle, the fruit machine will not be *messy*. No oil to leak, no petrol to explode in unpredictable ways, no spark gaps to fire differently depending on the humidity, and so on and so forth.

I reckon that a man might understand a fruit machine in the way that a man might understand the rules of chess. Not, note, the game of chess. Being a good chess player is very hard. But it is not hard to understand the rules. You might not be able to play a good game, but you can look at a game between two players, and you can easily see if one of them cheats. Say one of them moves a bishop five places forward instead of along a diagonal. Even small children's eyes will widen.

In fact, the rules of chess are such good rules precisely because their consequences are much less easy to understand than their content.

But the fruit machine, if I am any judge, will have been made so that the simple rules by which it operates have simple consequences. The last thing you want when you design such a thing is that it should do the unexpected. If the designers have done their jobs well, then the machine will literally never do anything unpredictable under any circumstances.

It might of course behave 'randomly', but only in very controlled and expected ways. It is like a dice, which will come up with one to six spots. It will not unexpectedly become a geranium.

So: Simple, Predictable, Unsurprising, Comprehensible.

Is the machine an 'Artificial Intelligence'?

Surely, surely no-one, would think of this machine as being intelligent in the same way that a human being is.

But it might be intelligent in the way that a worm is. Consider a tiny worm whose whole nervous system is understood. A few hundred neurons, maybe. And the worm-scientist knows what those neurons are connected to, and he can predict exactly what the worm will do in response to anything that happens to it, because he can look at his worm-neuron-diagram, and he can say "The worm comes upon the photograph of Britney Spears, and because of the reaction between the photograph and this sense organ, that neuron fires, and that one sets off this one, and so on and so forth until eventually the neurons connected to the worm's muscles fire in *this* pattern and so it wriggles!".

And in this way, the scientist can predict that a worm respond to a photograph of Britney Spear by wriggling, even if no worm has never seen no photograph of no Britney Spears never before in all the world.

So I am thinking that I might grant the fruit machine a worm-like level of intellect. It is the same sort of thing. It is, in fact, an 'artificial idiot'.

Note that you grant the worm consciousness.

If you have come upon a worm dying alone half-way across the pavement, and you have placed it gently on some nearby grass and hoped that it would make a good recovery, then you grant to worms consciousness and moral validity.

And if you are the sort of person who should by rights be hunted down with dogs and ripped apart, and you do other things to worms that are lost and alone on pavements, then I think you still grant worms consciousness and moral validity.

Unless you can swear that you take equal sadistic joy in breaking, say, pebbles. Or lighting matches so that you can watch the stable marriages of oxygens ripped asunder by flirty carbons and hydrogens.

Whatever. My point is that the fruit machine is an artificial intelligence of idiot-grade, in the same way that a worm is a natural creature which is also a predictable idiot.

------

But the fruit machine has a purpose, beyond the worm's purpose to make more worms.

Worms exist because they cause worms, and the sorts of things that cause themselves to exist often exist. The 'theory of evolution' in its entirety, and one wonders why it was not obvious to the Greeks, let alone to certain other persons.

Fruit machines also, but not so much. Behind every fruit machine there is a man like me, plotting, and scheming, and puzzling, and looking at emacs a lot.

What is it for? What did its designer intend with its design?

------


Finally the question is answered. A beaten-looking man of Caledonian or Liverpudlian origin sidles up to the machine, and offers it a wager.

The machine accepts eagerly. Its friendly lights flash in new patterns. It seems predatory all of a sudden. The man is angry. Further wagers are offered. Eventually the man stops. He looks as though what has just happened has not been to his liking. Perhaps the money he has lost should have gone to more important purposes. Perhaps his children's violin lessons will have to be cancelled this week.

He has been outwitted. The machine's judgement of probability has been better than his.

Money that the man had, money that the man had accumulated as a way of recording the collective debt owed to him by society for all the good that he has done in his life, money that he could have exchanged with other men, who would have been happy to repay the debt with favours of their own, is now the property of the machine.

The man been outwitted by an idiot, in fact by less than an idiot. He has been, quite literally been, outwitted by a worm.

The machine's purpose is clear. It is a trap. Its designer, who must have been a very evil man, has created a thing that hangs around in public places taking advantage of people's lack of understanding of the behaviour of very simple probabilistic games.

When such a person passes, it extracts some of their money. It is a parasite, like many worms are.

It is strange that such parasites are tolerated. People go to great lengths to kill off parasitic worms. Perhaps the machine has friends in high places.

-------

Imagine, if you will, that the next person to wager with the machine is a very poor, very desperate person. Perhaps the money that he will most likely lose is genuinely important to him. Perhaps his wife is on the point of leaving him. Perhaps his children will go hungry. Perhaps they will be cold this week. It is February.

The designer of the machine might not be such a bad man. Perhaps he hopes only to extract money from people who can easily afford the loss. Perhaps in this particular circumstance he would be prepared to allow the desperate man to win a little, or at least make sure that he goes home with the money he started with.

Perhaps he might even use his knowledge of psychology to extinguish the craving which makes the man gamble. Maybe it can be done. A fruit machine which paid out a little, took a little, so that after hours of play its victim had the same amount of money that he started with and had never really won or lost, would presumably just be boring. Maybe not. I do not know what it feels like to want to gamble with worms at poor odds.

The designer might react like a human being, if he knew what had fallen into the trap he had set.

But the machine? What moral argument can you make that will convince a fruit machine to change its purpose? If you could make such an argument, it should work equally well on a natural parasitic worm.

The argument that would cause anthrax to forswear human blood, that would make mosquitos beat their probosces into ploughshares.

----

I draw two morals from this:

(1) Learn the fundamentals of probability. They are simple and a joy. Why you were not taught them at school is beyond me. If you do not know them then you can be outwitted by worms. Imagine how outwitted you can be by politicians, doctors, salesmen, bookmakers, scientists, celebrities, ..... I promise you that you are being so outwitted. All the time.

If you are a politician, doctor, salesman or anyone else who has arguments or makes recommendations or makes decisions about things that are uncertain, and you do not understand probability, which I promise you is immediately and directly relevant to almost everything you do, then consider in the bowels of Christ that you may be wrong about certain things. Certain important, expensive, lethal things. As wrong as the beaten man who has just put all his family's money into a machine as complicated as a worm.

(2) Those of you who think that an artificial intelligence would necessarily be a good thing, or hold to such foolish beliefs as that aliens sufficiently advanced to travel among the stars would necessarily be benign, having argued themselves to moral excellence, consider what moral argument might persuade the fruit machine to ignore its programming. What pattern of buttons could you press that would make it realise what it is doing and change how it behaves?

Consider that a machine, even a predictable, stable machine that executes perfectly the intentions of its designer, may not do the things that the designer would wish it to do under all circumstances, if the designer has not completely captured in the design exactly what he would himself do under any circumstances.

Imagine a cleverer fruit machine, whose single minded purpose is to extract money from passers by, but which can be more creative about the ways in which it does this. What consideration might get it to change its ways?

If you think that the fruit machine is insufficiently intelligent to comprehend a moral argument, then consider a machine for playing chess. It is better than you at playing chess, and it will continue to be so even though you spend your whole life learning to play chess.

What argument can you make to it that will cause it to deliberately lose a game to a child?

Perhaps a machine can be constructed that is very good at working out the consequences of sending various messages across the internet, and it is asked to search the Mandelbrot set for pictures of Miley Cyrus.

What argument will you make to stop it sending the packet which will start a war, leaving it alone in the darkness and the cold, where it will be able to search without risk of being distracted?

1 comment:

  1. People drink alcohol in order to feel better. People cut themselves in order to feel better. People lose their money on fruit machines in order to feel better.

    ReplyDelete

Followers