As Nasim sat waiting for Blank Frank to be rebuilt, she suddenly had a vision of herself standing at the edge of an aquarium pool, trying to persuade a two-hundred-tonne blue whale to swat a ball through a hoop with its nose. However smart the animal was, and however agile it could be in the open ocean, the real trick was finding a way for it to move without crushing everything in sight.
She’d managed to shoehorn the project into Zendegi’s existing lease of computing resources from the Cloud; even so, the petabytes of data she was manipulating were almost choking the allocated racetrack memory. Paying for more storage wasn’t an option; the boss had agreed to let her play around for a while to see if this wild idea panned out, but only if she could keep the whole thing from showing up in the accounts.
The model-building algorithm she’d used all those years ago on the zebra finch data did not scale well, but after she and Bahador had spent a fortnight trying to refine it, to no avail, she’d decided to plough ahead regardless. They needed some results to show to investors as soon as possible; elegance and efficiency could come later.
Bahador knocked and entered the room; Nasim had taken to leaving the door half open to spare him the formalities.
‘We’re getting a strong demand surge from the south-east Asian arcades,’ he said. ‘We’re still below critical latency, but it’s tight.’
‘Okay.’ Nasim gritted her teeth. ‘If you have to, just… kill all my stuff.’
‘Will do.’ Bahador departed as quietly as he’d come.
You could have argued with me, Nasim thought irritably. At least gone through the motions, before accepting the inevitable. She watched the progress bars from her half-dozen tasks inching towards completion. Normally she would have been delighted that so many Indonesians and Malaysians were starting their weekend with an hour or two in Zendegi, but the racetrack drives she was sharing with these gamers were the only place she had to hold the intermediate results of her calculations. If Zendegi needed the storage there was no question of saving what she had in an offline back-up; writing that much data to holographic dye cubes would take hours. She’d simply have to hand the space over immediately, tossing a day’s work into the void.
The HCP had scanned the brains of more than four thousand subjects in unprecedented detail; most of the scans had been done on cadavers, but diffusion tensor imaging, which tracked the flow of water molecules along living nerve fibres, had also been employed. The full data set included equal numbers of males and females, but for the kind of composite model Nasim was trying to build it was worth analysing the sexes separately, removing at least one source of variation; the more alike the brains’ anatomy and organisation, the clearer the final picture would be. She’d started with the males, because she knew that if she’d chosen females for the demonstration everyone would have demanded to know why. But the model-building algorithm needed to generate temporary data for every possible pair of scans. Even using just the male subjects meant dealing with close on two million pairings.
One task reached its endpoint, saving its results and freeing up its working storage and processor allocation. Go see a movie instead, Nasim begged the teenagers of Kuala Lumpur. Just for tonight.
She brought up a latency histogram. It was flickering at the edge of critical, with a small proportion of customers experiencing minor, sporadic delays between their actions and the changes they wrought upon the virtual world. Mere head movements were handled locally, within the ghal’eha; no amount of congestion within Zendegi’s servers could disrupt the relationship between the user’s gaze and the image rendered in their goggles. But the object descriptions being fed to the castles needed to be updated rapidly enough to maintain the illusion of a fluid, responsive world. A tranquil stroll across a Martian desert might not suffer from a few extra milliseconds of latency, but a game of virtual table tennis could go downhill very fast. And while the brain was good at filtering out brief perceptual glitches, once they crossed a certain threshold all it could do was encourage you to stop indulging in risky behaviour for the sensorially confused – preferably after getting rid of the suspect contents of your stomach.
A second task finished. A third. Nasim peeked at one of the high-load games, looking down on a group of six hundred Indonesians who’d come together on a single, crowded battlefield to take on an army of leering demons. Along with their implausible rippling physiques, most of the good guys had magic charms and special powers – hard-won in gruelling quests, or stolen in battle, or maybe just bought with real money on the side. But nobody had signed up to play the spawn of the underworld; the enemy was entirely simulated. The game’s designers had the mechanics of a certain style of swordplay down pat, and Zendegi’s framework made the demons’ motions anatomically plausible, but aside from threats to tear out their opponents’ hearts they weren’t much good at repartee.
Perhaps none of the boisterous young men and women playing Minions of Eblis really wanted an opponent who was anything more than a robotic caricature of evil – and encouraging the beheading of more empathetic characters was not on any sane person’s wish-list. But in other games there were Proxies playing comrades and team-mates, guides and mentors, humble massed extras and world-shaking deities. Expanding the Proxies’ repertoire far beyond the range of these glowering, tongue-poking puppets would see players flock to Zendegi just as surely as if their competitors’ worlds had faded to black and white.
Nasim’s fourth task finished. It took all the self-discipline she could muster not to message Bahador to hold off regardless, to give her ten minutes more, no matter what havoc it caused. Even a few dozen customers with nausea or vertigo would be too high a price to pay; word of such incidents spread quickly, even when it was all down to idiots who couldn’t grasp the fact that they really were jogging after filling their stomachs with food.
The fifth task completed. Nasim glanced again at the Minions battlefield; the ground was soaked in green blood, the enemy almost vanquished, but if she remembered rightly, some of the demons had a habit of rising up and reconnecting their own heads. If she’d been feeling sufficiently ruthless she could have sought out the hidden levers that would keep these monsters down – but if news of her intervention ever got out, the stench would be worse than any number of ghal’eha full of vomit.
The sixth progress bar hit its endpoint and vanished. Nasim was dazed; she hadn’t expected her luck to hold out. Now that the neural model had been rebuilt and safely stored, the vast digital scratchpad needed to create it was no longer required. She could test the end product at her leisure.
Bahador appeared in the doorway and asked hopefully, ‘May I-?’ He must have been watching the race as closely as she had.
Nasim smiled. ‘Sure. Take a seat.’ Her desktop monitor had back-to-back screens; she switched on the flip-side and set it to mirror the main screen so Bahador could watch from his side of the desk.
Nasim took a few deep breaths. All the awkward manoeuvring with cranes and slings to get the whale into position for each trial took so much time and effort that she sometimes longed for a little more rigmarole surrounding the test itself. But she’d automated everything and now she could invoke the whole complex set-up with a single gesture. She pointed at the icon on the screen and mimed tapping it.
Half-a-dozen windows opened in rapid succession. The largest contained what looked at first glance like a real-time MRI image of a living human brain. A second glance revealed that an awful lot of vital brain tissue was missing, so it was hard to see how the subject could possibly be alive. Nevertheless, regions of the scan were lighting up, displaying patterns of normal, healthy activity that any neurologist would recognise.
A male voice began to speak in English:
‘It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness, it was the epoch of belief, it was the epoch of incredulity, it was the season of light, it was the season of darkness.’
The voice was faltering and flat. It might have belonged to an unenthusiastic, not-too-bright ten-year-old struggling to read aloud for his teacher, putting on a strange accent to amuse his classmates. The text-to-speech feature in the cheapest notepad was a model of perfect diction and clarity by comparison.
But this voice was not powered by a phonetic dictionary and a list of explicit contextual rules, assembled by lexicographers and linguists, then tweaked by comparison with millions of training sets. This voice was the product of nothing but the model-building algorithm’s best guess at the detailed neural wiring of half-a-dozen regions in a hypothetical adult human brain: a brain that was the functional best fit to the two thousand males that had been scanned for the HCP.
‘ “The epoch of incredulity”,’ Nasim echoed. ‘Not bad for a lowest-common-denominator vocabulary.’
Bahador smiled nervously. ‘Can you push it a bit further? See if it stays on track?’ The previous iteration had read as far as the second ‘times’ before it ceased paying attention to the text and began emitting a disturbing, not-quite-random word salad.
Nasim fed in another piece of the sample text. The simulation had no hands to reposition a virtual page, no head to tilt to change its angle of view, but it could move its virtual eyeballs across a paragraph, and it was predisposed to read whatever fell within its visual field.
‘It was the spring of hope,’ the voice continued. ‘It was the winter of despair. We had everything before us, we had nothing before us, we were all going direct to Heaven, we were all going direct the other way – in short, the period was so far like the present period, that some of its noisiest authorities insisted on its being received, for good or for evil, in the super… super… superlative degree of comparison only.’
‘One stutter,’ Nasim acknowledged. ‘After more than a hundred words. And all this with a skull full of empty space.’ She’d used a library of task-specific activation maps to select the regions of the brain to model – and in his present incarnation, Blank Frank possessed only the bare minimum needed to recite what was put in front of him. He had no ability to ponder the meaning of this quote from Dickens, to reflect on its imagery, to pursue its implications. He would not even recall the words a few seconds later; he possessed only working memory, a rolling present no deeper than it needed to be to parse a sentence. He could read a virtual teleprompter; anything more was beyond him.
As a proof of principle, though, he was spectacular. Every one of the two thousand male subjects scanned by the HCP would have acquired each word in their vocabulary under different circumstances. Over time, those words would have accreted vastly different associations. Hope? Despair? What kind of heart-breaking personal meaning had those strings of letters once carried for each of these men? Yet the algorithm had managed to peel away the tangle of idiosyncrasies and home in on the simplest common ground.
‘So how do you teach him Farsi?’ Bahador wondered. ‘Or Bahasa, or Arabic?’
‘He’s barely ten minutes old, and you want him to be a polyglot?’
‘Won’t he have to be, eventually?’ Bahador insisted gently.
Nasim rubbed her temples, half grateful, half annoyed to be brought down to Earth so quickly. How close were they to impressing a potential investor? Back in Redland’s lab her colleagues would have been awestruck, but winning over the Giorgio Omanis was another matter.
‘How do we persuade a mob of jaded algae-barons – who don’t even appreciate Zendegi as it is! – that a dim, monolingual zombie is the key to the next big thing in VR?’ She thought for a while. ‘Maybe we just need to find the right metaphor. Why not sell this as “strip-mining the brain”?’
Bahador looked uncomfortable. ‘Don’t forget that we’re talking about good Muslim boys; they might be partial to the occasional whisky, but any suggestion of grave-robbing is going to give them the heebie-jeebies.’
Nasim scowled. ‘Organ transplants are perfectly acceptable in Islam, so I really don’t see why-’
Bahador fixed her with an incredulous gaze. ‘You really think it’s that simple? I’m not saying that if some cleric was persuaded to consider the matter for a decade or two you couldn’t end up with a favourable ruling, but I wouldn’t take it as given that anyone else will see this as cut and dried the way you do.’
‘The donors all gave unrestricted consent,’ Nasim said stubbornly. ‘That ought to be the end of the matter.’
‘Fine. So if you want an investor, find one who thinks the same way. You must have met a few godless American businessmen who’d think nothing of “strip-mining” dead people’s brains.’
‘I was a student!’ Nasim retorted. ‘My friends were all students. Not everyone in America is a millionaire.’
Bahador spread his hands. ‘Well, don’t look at me. I’m the one who has to hide when the sheikhs come for tea.’ He glanced at his watch. ‘I’d better go keep an eye on the Indian surge if we don’t want to lose the last of our loyal fans there.’
Nasim nodded distractedly. When he’d left, she sat brooding. Personal connections weren’t everything, and it probably wouldn’t be too hard to get in touch with a few technology venture capitalists outside the company’s usual circle of Arab and Iranian backers. Anywhere in East Asia, Europe or North America her track record at MIT would at least get her a hearing. But even if she could find potential investors with no ethical or cultural objections to exploiting the HCP donors this way, how many of them would see Blank Frank as anything more than a novelty act? A digital mosaic of corpse brains that read Dickens would look about as promising to most people as a car engine based on Galvani’s twitching frog legs.
What she needed was someone at least as optimistic as she was about the technology’s potential. Someone who’d listen to Frank’s leaden recitation and picture a whole army of nimble, articulate Proxies rising from the same scans. Someone who’d convinced themselves long ago that the Human Connectome Project was destined to do more than help cure a few neurological diseases.
Maybe she should try the more conventional sources first, though, and leave Caplan as a last resort. She knew that he’d assembled a small empire of niche technology businesses, but she hadn’t really followed his fortunes; she wasn’t even sure that he still had money to burn. He could have squandered his inheritance on legal battles with the Superintelligence Project, trawling the world for homeless amnesiacs willing to swear that they were Zachary Churchland’s love-child.
But the longer Nasim spent turning the options over in her mind, the more it seemed that her reservations were just excuses to spare her pride. She couldn’t recall every word that she’d spoken to Caplan the last time they’d met, but she’d certainly spurned him comprehensively enough to make it awkward to come begging for his help all these years later.
Well, too bad. She couldn’t afford to shunt her best prospect to the end of the list. She had a choice between wasting six months collecting polite, frozen smiles from a dozen cautious entrepreneurs, or going straight to the loose cannon who’d once tried to force half a million dollars into her hands on the strength of a rabbit icon on a map of Cambridge.
Nasim entered the boardroom and took a pair of augmentation goggles from the cabinet near the door. She spent half a minute fussing with the strap, trying to make the goggles fit comfortably; the room was equipped with the technology for the sake of impressing clients, but she rarely had reason to use it herself.
She sat at the empty conference table and waited.
Caplan came online punctually, appearing to be seated directly opposite her. Their two tables were apparently close enough in size for the software to decide to re-scale everything so they were superimposed exactly, which made for less visual fuss. Nasim had deliberately cleared the space where she’d expected Caplan to show up, and he’d shown her the same courtesy, but elsewhere around the table chairs from the two locations intersected haphazardly. The room he was in was a little smaller than hers, so his walls were ghosted around her to remind her not to step through them.
‘You’re looking well, Nasim,’ Caplan said mildly.
‘You too.’ In fact, he barely seemed to have aged since they’d last met. Maybe the caloric restriction program was worth it after all, assuming this wasn’t just vanity and software. Nasim was showing herself honestly enough; routine distractions had kept her from updating her conferencing icon for about eighteen months, but she was quite sure that this small oversight was not enough to make her look twenty-five.
‘I read your proposal,’ Caplan said. ‘What you’ve achieved so far is impressive.’
‘Thank you.’ Nasim’s mouth was dry; she’d already grovelled a little in the email she’d sent him, but she was steeling herself for a certain amount of retributive gloating.
‘If I’ve understood you correctly,’ Caplan continued, ‘the main pay-off you see coming from this work is a set of stand-alone modules that your games developers will be able to call on, to assist them in programming a limited range of Proxy behaviours.’
‘That’s exactly right,’ Nasim replied. ‘There are some aspects of social – and even spatial – intelligence where humans still out-perform our best algorithms. If we can isolate a set of basic skills from the HCP data, then offer them up as a kind of library that developers can hook into with a minimum of fuss, that will benefit them enormously.’
‘But most aspects of the Proxies will remain the province of ordinary software?’ Caplan pressed her. ‘Biographical details, long-term context, strategic considerations-’
‘Of course,’ Nasim confirmed. ‘Even if we could construct a “whole brain” model that included the neural dynamics needed to support long-term memory, and even if we could figure out how to wire a Proxy’s notional back-story into that model… that wouldn’t just be wildly ambitious, it would be inefficient and highly inconvenient. Developers need to have a simple database of facts about their non-player characters that they can manipulate as easily as the geographical database for the game world. Sure, they want certain responses to the human players to be more natural, but not at the expense of an ability to follow predetermined scripts, or adopt explicitly programmed strategies. Trying to implement too much as part of the neural model would just make those things more difficult.’
‘Hmm.’ Caplan glanced down at some notes displayed on the table-top. ‘You mention reading players’ facial expressions?’
Nasim gestured at her goggles; they’d be invisible to Caplan, but they both knew they were there. ‘In less than a year, these will be replaced by contact lenses. It won’t be long before we routinely have an unobstructed view of players’ faces.’
Caplan said, ‘Sure, but don’t you think human face-reading skills have already been surpassed by micro-expression analysis?’
‘Perhaps,’ Nasim conceded. ‘But it might still be useful to developers to have something less forensic, for those times when they want a Proxy who seems attentive or sympathetic, rather than one who can beat you at poker or deliver a knock-out cross-examination in a murder trial. Anyway, that’s only one skill out of dozens.’
‘Many of the other skills are verbal, though, aren’t they?’ Caplan pointed out. ‘I’m still not quite clear what the deal is with language.’
‘Ah.’ Nasim knew this was her weakest point, but she confronted it head-on. ‘Obviously the HCP data itself will only encode knowledge of English; that’s the nature of the population they scanned. In the worst-case scenario, that could mean that we’d have to restrict our verbal modules to English, which we’d then license to VR providers with a large English-language customer base. Even so, that’s a huge market, so we could expect a decent return in terms of revenue even if we couldn’t use the same modules generally within Zendegi itself.’
‘And the less worse-case?’
‘We could probably exploit English-based modules in non-English scenarios to some degree, just by feeding their input and output through existing translation software – just as we do for players who don’t share a common language. But I’d hope we could do one better than that and train at least some of the modules to be bilingual.’
‘Expose them to huge data-sets and tweak them to give the right responses?’ Caplan suggested.
‘Yes.’
‘While hoping you don’t obliterate the very skills that make these networks more useful than the dumb translator software that gets created the same way?’
Nasim replied, ‘Well, yes. We can only try it and see how it pans out. What else should we do?’
Caplan said, ‘Have you ever heard of side-loading?’
‘No,’ Nasim confessed.
Caplan smiled slightly. ‘You didn’t research me very thoroughly, did you? Just checked that I still had a pulse and a bank balance.’
Nasim felt herself flush slightly, but took some comfort in the fact that her icon wouldn’t betray her response. She said, ‘I’m sorry if I didn’t dig deeply enough. I think I got distracted when I discovered that you owned an island in Cyber-Jahan.’
‘Doesn’t everyone? They’re so cheap.’ Caplan dismissed the matter with a wave of his hand. ‘Anyway… ten years ago I acquired a Swiss company called Eikonometrics.’ He slid a document across the table. Nasim glanced down; it was a report on the company’s research over the last fiscal year. A soft blue glow at the edges flagged it as a file Caplan had put into their shared dataspace, rather than a physical stack of paper.
‘They started out in subliminal image classification,’ Caplan explained. ‘They had a plan to harvest the brain-power of internet users by flashing up flowers and puppy dogs and road crash scenes just long enough to get a distinctive EEG signature. It was a very silly idea; some aspects of artificial vision might still be inferior to the equivalent human skills, but software caught up with the kind of results you could get on the cheap like that a long time ago.’
‘So why did you buy them?’
‘Their business plan sucked,’ Caplan said, ‘but they’d gained some genuine insight into neural information processing, and they’d started branching out into EBLD – Evidence-Based Lie Detection. Polygraphs are about as reliable as witch-dunking, so it’s not a huge boast to say that twenty-first century brain-scanning techniques have the potential to out-perform them. You stick a witness in a PET scanner or MRI machine and see what parts of their brain light up when they discuss their testimony or view certain images. There was a fad for that in the teens, and Eikonometrics got some grants and did some interesting research. But then it became clear that in any practical context there were too many problems interpreting the responses – and most jurisdictions started ruling it out on privacy as well as technical grounds. That’s when Eikonometrics shares hit rock-bottom, and I would have been crazy not to snap them up.’
‘Okay.’ Nasim was starting to get an inkling of where this was heading. ‘And side-loading…?’
‘Side-loading,’ Caplan replied, ‘is the process of training a neural network to mimic a particular organic brain, based on a rich set of non-intrusive scans of the brain in action. It’s midway between the extremes: in classic uploading, you look at the brain’s anatomy in microscopic detail and try to reproduce everything from that, whereas in classic neural-network training, what’s available to you is just stimulus and response: sensory input and visible behaviour, with the brain as a black box.
‘In side-loading, you get to peer inside the box, even if you don’t get to take it apart. You don’t have the resolution you’d get by peeling the brain with an ATLUM, but you have the advantage that you can expose the living brain to all kinds of stimuli – words, images, sounds, tastes, smells – and see how they bounce around inside the skull. And it doesn’t really matter how little external behaviour is evoked if you can watch the pattern of internal changes rippling out from every stone you toss in.’
Nasim said, ‘That all sounds fine in principle. But how far has it actually taken you?’
Caplan gestured at the Eikonometrics report. ‘Six months ago, we took a rat that had been trained to solve a particular maze. By observing its brain’s responses to a barrage of random sensory cues, we managed to modify an initially unrelated virtual rat to run the same maze.’
‘Observing its brain how? Do you mean micro-electrodes snuggling up to ten thousand neurons?’
Caplan shook his head. ‘Not at all. Purely non-invasive methods: multi-mode MRI and surface electrodes.’
Nasim was impressed now; this was more than she’d expected. ‘And with humans?’
‘With humans,’ Caplan said, ‘the problem is that we don’t really have a good generic virtual brain to use as a starting point. When the HCP published their results we started trying to construct one, but it looks as if you’re doing a much better job at that than we are so far.’
Nasim mulled over this frank assessment, sting included. The algorithm she’d used on the finch brains was public knowledge, but it wasn’t so cut and dried that someone coming in cold could apply it to the HCP scans and expect to crank out a good result. So she had a head-start on the path towards something Caplan wanted, badly – but it would be a mistake to overplay her hand and imagine that this made her indispensable.
‘So you think Eikonometrics can find a way to side-load languages into my Proxy modules?’ she asked.
‘I think it’s worth trying,’ Caplan said. ‘I think it’s your best chance. And there might be some other useful skills you could endow them with, the same way.’ He smiled. ‘Since I got your email, I’ve been giving that some thought. Suppose you could side-load motor skills from a few celebrity athletes: sign up the right soccer players and the Middle East is yours again. Add the right cricket players and Cyber-Jahan is history.’
Nasim was speechless. Motor skills were probably the simplest case imaginable – certainly less challenging than side-loading a second language. And Caplan was right: millions of people would vote with their credit cards to ‘play beside’ their sporting heroes, captured with an authenticity that no celebrity-endorsed console game could ever come close to achieving.
The most she’d been expecting from Caplan was a modest cash injection, giving her a chance to pursue her research far enough to see if it could be commercialised. Instead, he’d just offered Zendegi a plausible route to market domination.
Nasim forced herself to reply coolly, ‘It seems we both have something the other could use.’
Caplan inclined his head in agreement.
She said, ‘I’m going to have to discuss this with my superiors.’ If the Eikonometrics research was as promising as Caplan made it sound, Nasim was sure she could hook them on the sporting angle alone. ‘And there’ll be a complex joint venture agreement to be thrashed out-’
Caplan said, ‘I understand. Bring on the lawyers! But don’t run away and hide this time. Don’t forget, you’re at the heart of the deal.’
Nasim laughed, but she regarded his bland, youthful icon with unease. Caplan would emerge with a substantial stake in the company, but this wasn’t just about the cash flow that a synergistic marriage of their technologies could yield.
By setting up this meeting, she’d probably saved her job, her employer, her career… but no amount of commercial success with the Proxies would be enough for Caplan. She knew what the only worthwhile endpoint was for him – and she’d just agreed to harness herself to his cause.