White Death (19 page)

‘Maybe.’

‘All right. I’m here because Kwasi King, who half killed me this morning’ – he held up the arm with the cast on it – ‘is now on the run, suspected of three murders. We found e-mails from Mr Unzicker on his computer. I’ll come to that in a second. But what really interests me is this. You, Mr Nursultan, were at Madison Square Garden a few hours ago. The biggest star in your firmament is now making like O.J. Your championship match has gone down the can. Half the reporters in New York City must want to talk to you. I bet your sponsors sure do. You could spend your next forty-eight hours locked away in crisis meetings. And what do you do? You come up to MIT, which takes time, even if you
do
have a private jet, and you see a postgraduate student. Of all the things in the world you could be doing, this must be the most important to you. And if it’s the most important to you, I reckon it’s going to be pretty important to me.’

Nursultan regarded him levelly for a moment. ‘We work together. Unzicker and me.’

‘Work on what?’

‘Project.’

Patrese thought back to Unzicker’s e-mail, and took an educated guess. ‘Misha?’

He could almost hear Nursultan’s calculations: to lie, or not to lie? ‘Yes. Misha.’

‘And Misha is what?’

‘Misha nothing to do with you.’

‘Misha is something to do with Kwasi King, yes?’

‘Yes.’

‘Then it’s something to do with me. What is it?’

‘Can’t say.’

‘Can’t, or won’t?’

‘Very sensitive. Commercial. Very commercial sensitive.’

‘If I was any good at using things that were commercially sensitive, I wouldn’t be a Bureau agent. I’m not going to steal your idea or tell your competitors. But this is a direct murder investigation of Kwasi King, and if you don’t
co-operate, I will have you seven ways to Sunday on obstruc
tion of justice. That clear?’

Unzicker looked terrified. Nursultan weighed the odds. He’d tried to bribe Patrese before, asking him to manufacture a suspect in order to get Kwasi playing. Now Kwasi was a suspect himself, and he wasn’t playing. Nursultan had no leverage any more.

‘Misha not finished yet,’ he said.

‘And when it is?’

‘When it is … it first real artificial intelligence program in history.’

Unzicker explained the science to them.

When it comes to chess, he said, there are three main lines of computer application. The first was among the goals of the inaugural AI conference, held at Dartmouth – Ivy League again, Patrese thought – in 1956: to create a computer program which could defeat the world chess champion. At the time, that seemed a pipe dream, the very peak of human intellectual endeavor.

Forty years later, the first match between man and machine took place: Garry Kasparov, the world champion, against IBM’s Deep Blue. Kasparov won. A year later, Kasparov and Deep Blue played again. This time, Deep Blue won. The primacy of the machine was irrefutably established; so much so that nowadays there are several programs – all readily available in the shops for around $50 – that will get the better of any human on earth. Yes, even of Kwasi King. Not every time, not every game, but under anything approaching match conditions, a series alternating black and white, the machines will win. So far, so
Terminator
. No mileage in that for Unzicker.

The second line of application is the solution of chess itself; that is, with perfect play, is chess a win for White, a draw, or even a win for Black? Checkers was solved not so long ago; computers found a perfect line of play against which there is no defense. Effectively, checkers is over as a competitive sport. It’s like doing a crossword that someone’s already filled in.

But chess is a different matter altogether. Checkers pieces all move more or less the same. Not so in chess.
The differing moves of the pieces leads to a bewildering array
of possible positions: 10
¹²⁰
give or take a few. A thousand trillion trillion trillion trillion trillion trillion trillion trillion trillion trillion possible positions. To put that into
perspective, there’ve only been 10
²⁶
nanoseconds – not seconds,
nanoseconds
– since the universe was formed, and there are only around 10
⁷⁵
atoms in the entire universe. To all intents and purposes, chess is a road with no beginning and no end on which you could travel through eternity.

Yes, some positions have been solved: endgame positions with no more than six pieces (including both kings, who can never be taken) on the board. Supercomputers are now working on solving seven-piece endgames, and after that they’ll move on to eight, and so on. But each extra piece adds an exponential layer of complications and possibilities, not to mention the requirement for storage space. There are thirty-two pieces on a chessboard at the start of a match. At current rates of progress, it’ll be at least a couple of centuries before chess is solved, if indeed it ever is. No mileage for Unzicker here either.

It’s the third line of application, Unzicker said, that most interests him and Nursultan: the idea that chess is the perfect vehicle with which to create an intelligent computer program. A proper self-teaching, self-learning, self-changing program. One that would pass the Turing Test, the famous litmus test posed by Alan Turing, the father of modern computer science: that a device can be deemed intelligent if its answers to a set of human-posed questions are indistinguishable from those of a human.

As things stand, what makes computers so good at chess is precisely the fact that they
aren’t
human. They never get tired, flustered, hungry. They never worry about marriage problems or paying the bills. They never get upset by sudden noise or stale air. Their decisions are reliable, consistent and disciplined. They’re completely unemotional: a win, a draw, and a loss are simply three states, none of which affects their ability to play the next game. They have full knowledge of
everything programmed into them: openings, endgame posi
tions, middlegame strategies. They rely solely on logical inferences, and make no catastrophic errors: a computer never overlooks mate in one, either in its favor or against it. They evaluate 200 million positions per second, using complex algorithms to work out several different functions – placement, attack, blockade, exchange, skirmish, negation, and so on – crunching numbers with a brute force way beyond the furthest imagination of the keenest brain.

But they don’t have two things.

They don’t have intuition, gut feeling: that sixth sense that leads them to the right move through some process both inexplicable and primal.

And they can’t learn from their own mistakes. Whatever knowledge they have comes from new human input. When a computer makes a mistake, it will make that same mistake in the same conditions again and again until a programmer corrects it.

Why? Because the human brain and the computer chip are structured differently. The brain has what the chip does not: functional plasticity. In the human brain, memories are stored at a cellular level in the framework of the synapses. This framework can be altered, reconstructed, reconfigured, adapted to different functions. The brain’s plasticity is so great that it can even generate new specialized centers: for example, epileptics who lose the functions of one cerebral hemisphere can reorganize the other to take on the job of both. In contrast, a computer chip is literally hardwired, constrained by the rigidity of its own architecture.

What I’m doing, Unzicker said, is trying to combine the two. I’m growing neurons within a computer’s silicon circuits. And chess is the ideal proving ground for this. When it comes to exploring the cognitive processes, and developing problem-solving methods, chess has it all. There are six different types of pieces, all of which move in a different way. A pawn’s line of motion is different to his attack, a knight moves in three dimensions rather than two, a king and rook can perform a double move, the values of the pieces are constantly shifting, depending on their positions, and the king’s value is infinite. Strategy and tactics must combine in an endless dance. All these are priceless tools for reasoning. Goethe called chess the touchstone of the intellect.

And this, Unzicker said, is where Kwasi comes in. They’ve been constantly refining the project, experimenting, trying new things, failing, trying something else. At each turn, they’d needed Kwasi to play against the program and test its parameters. You want a program to think like a human, play chess like a human, with creativity and intuition? You have to test it against the best human around. Testing it against other chess programs would simply make it play like them.

Ever-increasing processing power and number-crunching wasn’t what the Dartmouth conference had had in mind all those years ago. Deep Blue and all its successors were intelligent only in the way a programmable alarm clock is intelligent. But that’s the way things are going now. Brute-force programs play the best chess. The market demands the best programs. So there’s no point bothering with anything other than brute-force programs. The competitive and commercial aspects of making computers play chess have taken precedence over using chess as a scientific proving ground.

But this one, Misha, will be different. I don’t care about computation, Unzicker said: I care about
understanding
. And with Nursultan providing backing, I don’t have to worry about running out of financial resources. As long as I keep working towards the goal, Nursultan will keep funding me.

Why Misha? In homage to the great Soviet player Mikhail Botvinnik. Botvinnik was world chess champion three times
after the Second World War: he was also an electrical engi
neer and computer scientist of great repute, trying to develop artificial intelligence not only in chess but also in economic management. His ideas were way ahead of the technology available at the time. This is their way – Unzicker’s, Kwasi’s and Nursultan’s way – of honoring him.

Patrese saw two things more or less at once.

First: this project would, if successful, revolutionize science. Proper, provable artificial intelligence was the Holy Grail: perhaps the final triumph of science over religion, the assertion of man’s primacy through the generation of life itself from inert matter. Whoever accomplished it would end up revered through the ages: a Newton, a Darwin, a Watson and Crick. That promise of immortality alone would explain Unzicker’s obsession with the project, and perhaps Nursultan’s too: whatever else you thought about him, the man wasn’t short of an ego.

But it was the second thing that really got Patrese thinking. He remembered reading somewhere that IBM’s share price had spiked 15 per cent in the immediate aftermath of Deep Blue beating Kasparov: a result which had, in essence, proved little other than that computers were very good at being computers. If Project Misha, potentially several magnitudes of achievement greater, were to come off, the effect on the Kazan Group – its share price, its market capitalization, its pretty much everything – would be astronomical.

‘What’s in it for you and Kwasi?’ Patrese asked Unzicker.

‘You mean money?’

‘I mean money.’

‘More than you can imagine,’ Nursultan said.

The manhunt for Kwasi had been going twenty-four hours, and had now spread across the entire north-east of the country. Police and sheriff departments in twelve separate states, plus federal law enforcement personnel, were all on the lookout for Kwasi, so far without joy. Not that there was a lack of information: quite the opposite. There seemed to be as many sightings of Kwasi as there were of Elvis, and they were all just as unreliable. The task force was receiving hundreds of calls an hour, and even with every available person dragooned into service – not only Bureau agents and cops, but administrative support staff, press officers, guys who worked in HR and IT, pretty much anyone who could answer a phone and take down basic details – they were still overwhelmed. Someone rang to tell them that they’d never find Kwasi because he’d been abducted by aliens; another said the murders were connected with the Kennedy assassination. Crazy Kwasi, the papers were calling him; but he was no more crazier than some of the folks out there, Patrese thought.

Patrese was always wary of letting facts fit theory rather than vice versa, but at every turn he was becoming more convinced that Kwasi was indeed playing some murderous chess game. For example, he’d realized while out running in New Haven early this morning how closely the layout of the central streets mirrored that of a chessboard. The Green, where the first two bodies had been found, was the middle of nine squares, each one delineated by street boundaries.

In chess, a king can move one square in any direction. If a king – Kwasi King? – was standing on the Green, he could move to any one of those eight adjacent squares, those eight city blocks. And if this was a chess game, who better to ask advice from than the second-best player in the world? Well, the best player in the world, perhaps, but he was currently indisposed. Which was why Patrese had come to see Rainer Tartu in New York.

Like Nursultan, Tartu was staying at the Waldorf-Astoria. Unlike Nursultan, he didn’t have a suite all to himself, though his room was still hardly tenement standard. He was packing when Patrese arrived.

‘You leaving already?’ Patrese asked.

‘I have no reason to stay here. The match is indefinitely postponed.’

‘Going back to Europe?’

Tartu shook his head. ‘To New Haven.’

‘New Haven, Connecticut?’

‘That’s right.’

‘That’s where I am. That’s where we’re based during this …’

‘Then maybe I’ll see you around. It’s not such a big place, I hear.’

‘What are you doing there?’

‘I have three passions, Mr Patrese. Chess is the first, of course. I’ll be playing a few simultaneous exhibitions up around there.’