Clavain smashed the ice again, dislodging another layer.
‘You…’ Setterholm paused. ‘You could bring the whole place down on us if you’re not careful.’
Clavain raised the pick again, letting out a groan of effort as he swung. This time he put all his weight behind the swing, all his fury, and a chunk the size of his upper body calved noisily from the wall.
‘I’ll take that risk,’ Clavain said.
‘No. You’ve got to stop.’
‘Why? It’s only ice.’
‘No!’
Setterholm rushed him, knocking him off his feet. The ice pick spun from his hand and the two of them crashed into the ground, Setterholm landing on his chest. He pressed his faceplate close to Clavain’s, every bead of sweat on his forehead gleaming like a precise little jewel.
‘I told you to stop.’
Clavain found it difficult to speak with the pressure on his chest, but forced out the words with effort. ‘I think we can dispense with the charade that you’re Iverson now, can’t we?’
‘You shouldn’t have harmed it.’
‘No… and neither should the others, eh? But they needed that ice very badly.’
Now Setterholm’s voice held a tone of dull resignation. ‘For the reactor, you mean?’
‘Yes. The fusion plant.’ Clavain allowed himself to feel some small satisfaction before adding, ‘Actually, it was Galiana who made the connection, not me. That the reactor ran on ice, I mean. And after all the outlying bases had been evacuated, they had to keep everyone alive back at the main one. And that meant more load on the reactor. Which meant it needed more ice, of which there was hardly a shortage in the immediate vicinity.’
‘But they couldn’t be allowed to harvest the ice. Not after what I’d discovered.’
Clavain nodded, observing that the reversion from Iverson to Setterholm was now complete.
‘No. The ice is precious, isn’t it? Infinitely more so than anyone else realised. Without that ice the worms would have died—’
‘You don’t understand either, do you?’
Clavain swallowed. ‘I think I understand more than the others, Setterholm. You realised that the worms —’
‘It wasn’t the damned worms!’ He had shouted — Setterholm had turned on a loudspeaker function in his suit that Clavain had not yet located — and for a moment the words crashed around the great ice chamber, threatening to start the tiny chain reaction of fractures that would collapse the whole structure. But when silence had returned — disturbed only by the rasp of Clavain’s breathing — nothing had changed.
‘It wasn’t the worms?’
‘No.’ Setterholm was calmer now, as if the point had been made. ‘No — not really. They were important, yes — but only as low-level elements in a much more complex system. Don’t you understand?’
Clavain strove for honesty. ‘I never really understood what it was that fascinated you about them. They seem quite simple to me.’
Setterholm removed his weight from Clavain and rose up onto his feet again. ‘That’s because they are. A child could grasp the biology of a single ice-worm in an afternoon. Felka did, in fact. Oh, she’s wonderful, Nevil.’ Setterholm’s teeth flashed a smile that chilled Clavain. ‘The things she could unravel… she isn’t a failure; not at all. I think she’s something miraculous we barely comprehend.’
‘Unlike the worms.’
‘Yes. They’re like clockwork toys, programmed with a few simple rules.’ Setterholm stooped down and grabbed the ice pick for himself. ‘They always respond in exactly the same way to the same input stimulus. And the kinds of stimuli they respond to are simple in the extreme: a few gradations of temperature; a few biochemical cues picked up from the ice itself. But the emergent properties…’
Clavain forced himself to a sitting position. ‘There’s that word again.’
‘It’s the network, Nevil. The system of tunnels the worms dig through the ice. Don’t you understand? That’s where the real complexity lies. That’s what I was always more interested in. Of course, it took me years to see it for what it is—’
‘Which is?’
‘A self-evolving network. One that has the capacity to adapt; to learn.’
‘It’s just a series of channels bored through ice, Setterholm.’
‘No. It’s infinitely more than that.’ The man craned his neck as far as the architecture of his suit would allow, revelling in the palatial beauty of the chamber. ‘There are two essential elements in any neural network, Nevil. Connections and nodes are necessary, but not enough. The connections must be capable of being weighted; adjusted in strength according to usefulness. And the nodes must be capable of processing the inputs from the connections in a deterministic manner, like logic gates.’ He gestured around the chamber. ‘Here, there is no absolutely sharp distinction between the connections and the nodes, but the essences remain. The worms lay down secretions when they travel, and those secretions determine how other worms make use of the same channels; whether they utilise one route or another. There are many determining factors — the sexes of the worms, the seasons, others I won’t bore you with. But the point is simple. The secretions — and the effect they have on the worms — mean that the topology of the network is governed by subtle emergent principles. And the breeding tangles function as logic gates; processing the inputs from their connecting nodes according to the rules of worm sex, caste and hierarchy. It’s messy, slow and biological — but the end result is that the worm colony as a whole functions as a neural network. It’s a program that the worms themselves are running, even though any given worm hasn’t a clue that it’s a part of a larger whole.’
Clavain absorbed all that and thought carefully before asking the question that occurred to him. ‘How does it change?’
‘Slowly,’ Setterholm said. ‘Sometimes routes fall into disuse because the secretions inhibit other worms from using them. Gradually, the glacier seals them shut. At the same time other cracks open by chance — the glacier’s own fracturing imposes a constant chaotic background on the network — or the worms bore new holes. Seen in slow motion — our time frame — almost nothing ever seems to happen, let alone change. But imagine speeding things up, Nevil. Imagine if we could see the way the network has changed over the last century, or the last thousand years… imagine what we might find. A constantly evolving loom of connections, shifting and changing eternally. Now — does that remind you of anything?’
Clavain answered in the only way that he knew would satisfy Setterholm. ‘A mind, I suppose. A newborn one, still forging neural connections.’
‘Yes. Oh, you’d doubtless like to point out that the network is isolated, so it can’t be responding to stimuli beyond itself — but we can’t know that for certain. A season is like a heartbeat here, Nevil! What we think of as geologically slow processes — a glacier cracking, two glaciers colliding — those events could be as forceful as caresses and sounds to a blind child.’ He paused and glanced at the screen in the back of the imaging radar. ‘That’s what I wanted to find out. A century ago, I was able to study the network for a handful of decades, and I found something that astonished me. The colony moves, reshapes itself constantly, as the glacier shifts and breaks up. But no matter how radically the network changes its periphery, no matter how thoroughly the loom evolves, there are deep structures inside the network that are always preserved.’ Setterholm’s finger traced the red mass at the heart of the green tunnel map. ‘In the language of network topology, the tunnel system is scale-free rather than exponential. It’s the hallmark of a highly organised network with a few rather specialised processing centres — hubs, if you like. This is one. I believe its function is to cause the whole network to move away from a widening fracture in the glacier. It would take me much more than a century to find out for sure, although everything I’ve seen here confirms what I originally thought. I mapped other structures in other colonies, too. They can be huge, spread across cubic kilometres of ice. But they always persist. Don’t you see what that means? The network has begun to develop specialised areas of function. It’s begun to process information, Nevil. It’s begun to creep its way towards thought.’
Clavain looked around him once more, trying to see the chamber in the new light that Setterholm had revealed. Think not of the worms as entities in their own right, he thought, but as electrical signals, ghosting along synaptic pathways in a neural network made of solid ice…
He shivered. It was the only appropriate response.
‘Even if the network processes information… there’s no reason to think it could ever become
