Rana wouldn’t lift her gaze from the ground, and she resisted Martin’s attempts to distract her with small-talk and optimistic prognoses. He was trying to keep his own imagination in check; he knew what went on in Evin Prison, but nobody was going to round up and torture every last Iranian who’d ever stocked contraband action movies. Only if they’d traced Shokouh’s false passport back to Omar would he be in real danger.
Martin spotted a woman further along the queue speaking on a phone, though she was doing her best to hide it in her sleeve. As far as he knew the Slightly Smart phones weren’t illegal, though perhaps they soon would be.
When she hung up the call, she turned and spoke agitatedly with her neighbour. Whatever the subject, it was not a private matter; within minutes Martin could see the news being spread up and down the line. Maybe the authorities had decided to charge Jabari after all; if his resignation hadn’t been enough to win back conservative support, why not pull out all the stops and have a show trial, to prove that nobody was above the law?
But any mention of Jabari always conjured up at least a few wry smiles. Nobody was smiling as they heard this news.
The rumour finally reached Mohsen and Rana; Martin’s Farsi had largely deserted him, but once he had heard Ansari’s name mentioned he could think of only two possibilities.
‘Have they arrested him?’ he asked.
‘No,’ Rana said, ‘he’s been shot. They’ve taken him to hospital, but he’s not expected to last the night.’
6
Nasim hunched over her computer screen, gazing intently at a section of code from her neural map integration routines, blocking out thoughts of anything else.
No two zebra finches sang exactly the same song; no two finches had identical brains. So how could you use partial, imperfect images of a thousand different finch brains to build up some kind of meaningful composite?
On a gross level the same structures within the brain appeared in more or less the same anatomical locations, but as you zoomed in towards the level of individual neurons, the cues that counted most were the cells’ biochemistry and their patterns of connections. The problem lay in keeping the notion of a pattern of connections from becoming meaninglessly vague, uselessly rigid, or maddeningly circular. If ten thousand cells of biochemical type A sent axons to ten thousand cells of type B, that certainly didn’t mean that they were all interchangeable. But if you insisted that only neurons that were wired up to identical neighbours in identical ways could be treated as common features, there would be no matches at all. Worse, if you could only characterise every neuron by first characterising the neurons to which it was joined, you ran the risk of pushing everything down a rabbit hole of endless self-reference. The whole endeavour was like trying to reconstruct the human skeleton from a thousand incomplete – and partly inconsistent – translations of ‘Dem Dry Bones’ into unknown foreign languages. ‘The fifflezerm’s connected to the girglesprig…’
Over the months she’d spent working on the problem, Nasim had tried all manner of high-powered statistical techniques and classification schemes from abstract network topology, but the approach that was finally showing signs of a payoff involved searching for distinctive sub-networks, not by their pattern of connections per se, but by their function. An engineer staring at a circuit diagram could group the components into various kinds of functional blocks – say, half-a-dozen that formed an oscillator, another half-dozen comprising a filter – without requiring an absolutely rigid, unvarying design for each of these meta-components. An oscillator was anything that oscillated; it didn’t have to be a perfect match for the first one you’d encountered in a textbook. Similarly, if a group of neurons had the same general effect on their inputs as another group, it didn’t really matter if there happened to be thirty-nine neurons in one group and forty-five in the other. ‘The same general effect’ was easier said than defined, but Nasim had been refining the notion for weeks now, and she was convinced that she was finally closing in on a set of meaningful categories.
She tweaked a few definitions in her code and started it running again. It would take a couple of minutes to process the full data set; she looked away from her screen and across the lab. Everyone was unnaturally quiet today; Redland was down in Washington, testifying before a House Select Committee on the mooted Human Connectome Project, and Judith had gone with him. The Committee had been holding hearings for a month, and Redland was just one of dozens of scientists who’d been called to give testimony, but the occasion of his trip had reminded everyone that their funding, and their future, lay in the balance.
The composite map appeared on the screen. Nasim was about to slip on her headphones when a mischievous impulse took hold of her. She pulled the headphone plug out of its jack, rerouting the computer’s audio to its speakers. Then she fired up the software syrinx and ran the latest simulation of the finch brain’s vocalisation pathways.
The infantile babbling of her early trials had slowly been giving way to a more ordered song, but this time hairs rose on the back of her neck. The distinctive rhythms of an adult bird’s call – the whole style, the whole structure – were finally present.
With the song still playing, she checked the simulation’s virtual EEG. The waveforms were not an exact match to any of the biological recordings on file, but the statistics all fell within the population ranges. If she’d handed the traces to a neurobiologist, they would not have been able to pick the artificial one from the real.
Mike stepped away from his bench and looked around, annoyed. ‘Who took the bird out of the animal house?’ he demanded. He was wearing a hairnet and something that resembled a plastic shower cap. ‘If I get droppings in my cell cultures, that’s a month’s work down the tube!’ He finally homed in on the sound and turned to glare angrily at Nasim. ‘Where is it?’
It took her a moment to realise that he wasn’t joking. She said, ‘No droppings, Mike, I promise.’
Mike, Shen and Dinesh gathered around her desk and watched as she ran through a battery of further tests. She kept the syrinx warbling, trying to shake off the eerie feeling that she’d stitched together something gruesome from the corpses of the birds and could now feel the awakened result fluttering its wings in her hand.
Shen said, ‘We should play this to a female bird and see if she’s attracted. A Turing test for zebra finches.’
‘No,’ Mike countered, ‘we should simulate a female’s auditory centres, and see if that simulation is attracted.’
‘One program fools another program? How is that a test?’ Shen demanded.
‘It’s not a test,’ Mike agreed, ‘but it would be much easier for them to consummate the relationship.’
Shen pondered this. ‘I think the Media Lab could put together some avian tele-dildonics faster than we could construct a purely software female capable of mating.’
‘Can we cut the Bride of Frankenfinch crap?’ Nasim pleaded. ‘There’s nothing in there but the vocalisation PDP. If that can feel lust all by itself, then so can a Casio keyboard.’
Dinesh said, ‘There’s nothing in there that can feel lust, yet. But now that you can integrate maps from different imaging techniques, it would take us, what, eighteen months to do the whole finch brain?’