Quarantine

by Greg Egan

  'All / want,' he says, 'is for everyone to be satisfied. It's as simple as that.'

  Three days after the meeting, I take a small detour on my way home from the underground. I drop in at a stall which sells downmarket consumer pharmaceuticals and nano-ware: smart cosmetics, active tattoos, 'natural' sex aids (meaning, they act on nerves in the genitals, not the brain), muscle 'enhancements' (painless short cuts to dysfunctional hypertrophy), and the kind of neural mods that belong in cereal packets. I don't know which backstreet manufacturer Lui employed to create his collapse-inhibiting mod, but collecting the finished product from a place like this doesn't exactly fill me with confidence.

  I quote the order number Lui gave me, and the stall owner hands me a small plastic vial.

  Before going to bed, I spray the vial's contents into my right nostril, and a heavily modified version of Endamoeba histolytica -the protozoans responsible for amoebic meningitis, amongst other delights - carry their burden of nanomachines into my brain. I lie awake for a while, thinking about the daunting navigational and constructional feats that the virus-sized robots are expected to perform - and wishing I'd asked Lui just how much experience he's had with mod design. For all I

  153

  know, the manufacturers might have used the most reliable, modern hardware available to build and program the things - but even perfectly constructed nano-machines can do perfectly fatal damage, if they're following a design that turns vital brain centres into neural spaghetti.

  Eventually I give up worrying. I'm doing all I can to serve the true Ensemble, and if I can't find peace in that alone . . .

  I stare up at the ceiling, at a thin strip of morning sunlight breaking in through a crack in the blinds. I choose sleep.

  Boss wakes me three hours early, as requested. Well, I'm not dead, paralysed, deaf, dumb or blind. Yet. I run integrity checks on all my other mods, and none have been damaged - but then, that's the least likely mistake of all. Neurons that are already part of existing mods are tagged with cell-surface proteins which no correctly functioning nanomachine could miss - and are also altered in other ways which would need to be deliberately reversed before they could be stimulated into changing their synaptic connections.

  Lui gave me no name to invoke, so I have MindTools (Axon, $249) perform an inventory; it can't 'scan' my whole skull by any means, but it can send a standard 'announce yourself request down the inter-mod neural bus, and list the replies it gets back. Only the loyalty mod remains silent, refusing to name itself, or even to admit its presence.

  The collapse-inhibiting mod turns out to be camouflaged, hidden inside a cheap-and-nasty games mod called Hypernova (Virtual Arcade, $99). Hypernova is to von Neumann what, in my childhood, a dedicated games machine was to a personal computer. I flip through its menus and help text. It can be loaded with software from ROMs or on-line libraries, either through an IR mod like RedNet, or the crude, old-fashioned way: modulated visible light.

  154

  I might as well make the camouflage plausible; nobody has a games mod with nothing in it. I phone Virtual Arcade's library. The current best-seller is an historical war game for brain-dead weapons fetishists called Basra 91, boasting authentic missile's-eye views of the genocide. I pass on that, and download last week's favourite, Metachess. 'Every configuration of pieces generates a unique set of rules.'

  I play the game for a while (losing badly on novice level), trying to invoke all of the mod's facilities in turn, but after twenty minutes I still haven't found the trapdoor into the real thing. I'm beginning to wonder if some elaborate sequence of commands is necessary, when I realize that there's still one function that I haven't touched. I go back to the downloading menu and invoke the archaic visible light option. Instead of receiving the expected complaint - that I'm not staring at an appropriate data source - a new menu appears, bearing only two words: OFF and ON. There's a tick mark beside OFF.

  I hesitate, but the fucking thing has to be tested, sooner or later - and if it's going to malfunction horribly, I'd rather find out about it here and now than in the anteroom of Po-kwai's apartment.

  The distinction between idle visualization and an active command to a mod is hard to describe - but it's as easily mastered, and forgotten, as the difference between real and imagined actions of the body. Only under stress does it cease to feel like second nature. As I picture the tick mark reappearing beside the word ON, I'm acutely aware of the fact that the mental image I'm manipulating is the menu itself.

  Nothing happens, nothing changes - which is exactly as it should be. I hold my hand up before my eyes, and it conspicuously fails to dissolve into a blur of alternatives. The whole room remains as solid, as ordinary, as ever. So far as I can judge, my mental state is entirely unaltered -except for a predictable surge of relief to find that I'm still not paralysed, blind or detectably insane. Lui might have

  155

  known what he was doing, after all. The mod might even be working.

  In which case, I am now smeared - even if there are no observable consequences whatsoever. The uniqueness, the solidity, the utter normalcy of everything, is a product of the fact that I will be collapsed at some time in the future - this time, without Po-kwai's eigenstate mod to distort the probabilities, or to mix and confuse the alternatives.

  I will be collapsed? Perhaps it makes more sense to assume that I'm 'already' being collapsed - at a time which only seems to lie in the future - and this whole experience is arising 'retrospectively' from that process. When the spin of an ion is measured, Po-kwai assured me, that is when it becomes definite, not before.

  I laugh out loud. In spite of everything - Laura's feats of escapology, Po-kwai's success with the ions, my own impossible mod failures - it's still not real to me. And in spite of the fact that I know that this is the heart of the true Ensemble ... it still sounds like a load of pretentious, inconsequential, undergraduate philosophical crap. For all I know, I've just installed the Emperor's new mod.

  I bring back the menu, tick the OFF switch -

  - and wonder: what about all of the versions of me who didn't just do that? Have they been destroyed by the wave-collapsing pathways in my skull . . . even though half of them may have been scattered around the room -or across the city - by now?

  They must have been - destroyed by me, or destroyed by some other observer.

  All of them?

  Forget the collapse-inhibiting mod - that changes nothing but the timing. The ordinary course of events must add up to normality. However frequently or infrequently the brain performs the collapse, it must reach out and destroy even the most far-flung, improbable states. If not, then these untouched states would persist indefinitely. There's no point appealing to other observers to clean things up; they'd do the job

  1-56

  imperfectly, too. If the collapse were not all-consuming, then the single, solid branch of reality wouldn't be unique at all. It would lie in the centre of a huge void of depleted alternatives, but that void would be finite - and beyond it would lie an infinite thicket of fine branches, the ghosts of improbabilities too remote to have been destroyed. And that's just not the way things are.

  I start my own experiments while Po-kwai is still waiting for the next phase of her work to begin. Perhaps that's pointless, given that - so far - the most dramatic effects have occurred on those nights when she's actually used the eigenstate mod successfully. But I can't see the harm in trying - and I might as well be optimistic. If my own use of the eigenstate mod remains tied absolutely to hers, it could end up taking me years to achieve the simplest tricks - let alone any massively improbable cross-town burglaries.

  Po-kwai developed her skills working with the simplest possible systems: silver ions carefully prepared to consist of an equal mixture of just two states. I don't have access to anything so pure, but I can still work on the same basic principle: taking a system which would normally collapse according to well-known probabilities, and trying to skew the odds. Both von Neumann and Hy
pernova have facilities for true random-number generation - as opposed to the deterministic pseudo-random sequences produced by purely algorithmic means. They employ groups of neurons specially tailored for the purpose, balanced on a fractal knife-edge between firing and not firing, stuttering chaotically in the sway of nothing but intracellular chemical fluctuations and, ultimately, thermal noise. Ordinarily, the system should collapse in such a way as to generate random numbers spread uniformly throughout a specified range; any skew, any bias, would mean that I'd succeeded in changing the probabilities - favouring one of the system's states to make it more likely to be the sole survivor of the collapse -just as Po-kwai succeeded in increasing the probability of the up state in her silver ions.

  157

  I spend three nights trying to influence von Neumann's random numbers, with no success . . . which is no great surprise. The combination of visualization and wishful thinking I employ - for want of anything better - seems more like an exercise for aspiring psychics than an attempt to give a precise command to a specific neural mod, whoever's skull it happens to be in. Lui is no help; he's never so much as caught a glimpse of a description of the eigenstate mod's interface. So, I laboriously steer a conversation with Po-kwai onto the topic (and probably succeed in sounding far less natural than if I'd just asked her, out of the blue).

  She says, 'I've told you: I don't remember using that part of the mod; I just switch on the collapse-inhibitor, then sit back and watch the ions. The two functions are independent. The whole thing was installed as a single package, but in effect, it's two separate mods. The eigenstate mod only works when it's smeared . . . and while I'm smeared, I can - evidently - operate this smeared mod. After the collapse, though, I know nothing about it.'

  'But. . . how can you have learnt to do something that you don't even remember doing?'

  'Not all skills rely on episodic memory. Do you remember learning how to walk? Sure, if I've grown better at manipulating eigenstates, then that skill must be embodied in some kind of neural structure, somewhere in my brain - but certainly not as a conventional memory, and probably not in any form which could ever make sense to me, or be of use to me, while I'm collapsed. I mean, the eigenstate mod is a neural system which only works when it's smeared, so there's no reason why other parts of my brain - pathways which formed naturally, during the course of the experiment - might not also work only when smeared.'

  'You're saying that when you're smeared, you know how to work the eigenstate mod - but the knowledge is encoded in your brain in a way that's unreadable when you're collapsed?'

  158

  'Exactly. The knowledge must have been stored in the brain while I was smeared ... so it's hardly surprising that I can only decipher it when I'm smeared again.'

  'But . . . how can information about being smeared survive from one time to the next, when the collapse wipes out every last trace of every eigenstate but one?'

  'Because it doesn't! That's only true if the eigenstates don't have a chance to interact - and the eigenstate mod means they do interact. There's nothing new, in principle, about smeared systems leaving proof that they've been smeared; half the critical experiments in early quantum mechanics relied on it. Indelible evidence of multiple states co-existing is more than a century old: electron diffraction patterns, holograms . . . any kind of interference effect. You know, the old photographic holograms were made by splitting a laser beam in two, bouncing one beam off the object, then recombining the beams and photographing the interference pattern.'

  'What's that got to do with smearing?'

  'How do you split a laser beam in two? You point it at a sheet of glass with a very thin coating of silver, angled at forty-five degrees to the beam; half the light is reflected to the side, while the rest passes straight through. But when I say "half the light is reflected," I don't mean every second photon is reflected - I mean every individual photon is smeared into an equal mixture of a state where it's reflected, and a state where it passes straight through.

  'And if you try to observe which path each photon takes, you collapse the system into a single state - and you destroy the interference pattern, you ruin the hologram. But if you let the beams recombine, unmolested, giving the two states a chance to interact, then the hologram remains as tangible, lasting proof that both states existed simultaneously.

  'So, maybe interactions between different versions of my brain can leave some kind of permanent record of the experience of being smeared. And just as a laser-light hologram is an indecipherable mess to the naked eye -bearing no resemblance to the object whatsoever, until

  159

  the image is reconstructed - this information stored in my brain may be incomprehensible to me, but presumably it comprises skills that are useful to the smeared Po-kwai.'

  I digest this. Okay. But even if there is this way for "the smeared Po-kwai" to learn things that you don't know about . . . what did you actually do to encourage her to learn what you wanted her to learn?'

  'Chanting the ion deflections may have helped. But I suspect that just wanting the experiment to work, badly enough, was all it took. The more I wanted it, the greater the number of versions of me who'd still want it, once I was smeared - and so the total smeared Po-kwai must have ended up wanting it, too. Anything else would have been highly undemocratic' She says this tongue-in-cheek, but not entirely.

  I say, 'At last - a rigorous definition of seriousness of purpose: when you diverge into multiple selves, how many stick to your stated goal, and how many abandon it?'

  Po-kwai laughs. 'Sure. You could quantify anything at all that way. How do I love thee? Let me count the eigenstates . . .'

  At home, deprimed, I wonder about my own goals, my own seriousness of purpose. Nothing that happened on the two occasions when I was (noticeably, memorably) smeared was anything that / wanted. And now? / may fervently wish to serve the true Ensemble by learning to steal the eigenstate mod - but once I'm smeared, how does the voting go?

  I've never deluded myself: I've never pretended for a moment that I'd be the same without the loyalty mod. But from what Po-kwai has told me about the meaning of the wave function, I'd have assumed that the very fact that the loyalty mod works, reliably, must reflect a high probability for those quantum states in which it keeps on working. Smearing may create some versions of me for whom the loyalty mod has failed - but they ought to be massively outnumbered by versions for whom it still functions.

  160

  And yet ... I deprimed with P3 still running; I saw Karen without invoking her. In both cases, the same argument should apply: the majority should have been backing the status quo. But the status quo was not maintained.

  So what exactly is going on when I smear in the anteroom and try - or think I try - to sway the random numbers being spat out by von Neumann? Nothing of consequence ... or a virtual war between a billion possible versions of who I might become? Pitched battles for the eigenstate mod, the super-weapon, the reality shaper? All I end up knowing about is the subsequent stalemate - but maybe the balance of power is gradually shifting, maybe there are 'holograms' in my head which record the changing state of play.

  The thought that there might be versions of me coming into being who act against my wishes, who fight against everything I'm living for, is so repugnant that all I want to do is mock it, dismiss it as absurd. And even if it is true . . . what can I do about it? How can I make a difference to the outcome of these battles? How can I reinforce the factions which remain in the grip of the loyalty mod -which remain loyal to me?

  I have no idea.

  I give up on von Neumann; there's something highly dubious about aiming to influence neurons in my own skull. In a junk market close to my building, I find an electronic dice generator, about the size of a small playing card. The heart of the device is a tiny sealed unit containing a few micrograms of a positron-emitting isotope, surrounded by two concentric spherical arrays of detector crystals. This set-up is immune - the seller's know-it-all hologrammic
spruiker assures me - to both natural background radiation and any deliberate attempts to tamper; no external event can be confused with the characteristic pair of gamma rays produced when a positron is annihilated within the device itself. 'Of course, if the gentleman would prefer a model more amenable to discreet persuasion . . .'

  161

  I buy the tamper-proof version. The software can produce any desired combination of polyhedra; I select the traditional pair of cubes, and spend an hour testing the thing. There's no trace of bias.

  I take it with me on duty, and when Po-kwai is asleep, I sit in the anteroom, deprimed, smeared and collapsed by Hypernova, trying to imbue my virtual selves with a sense of purpose that might survive the wave function's inexorable dispersion. I feel a twinge of guilt about intentionally depriming, abandoning my responsibility to Po-kwai, but I can't risk having P3 interfere with the collapse in unpredictable ways. And I tell myself: if the Children ever do find out that ASR is engaged in blasphemous research, they'll simply bomb the building, and there'll be nothing I can do about it, primed or not.

  The dice remain scrupulously fair.

  Po-kwai begins the third phase, another measurement of correlations within her brain. I can understand Lui's impatience with these inward-looking experiments - but at the same time, I can appreciate, more than ever, ASR's reasons for proceeding cautiously. I may know for a fact that all kinds of macroscopic feats are possible, but I'm thrashing around in the dark trying to master them, and taking huge risks in the process. Left to themselves, ASR might take ten years before they try anything similar - but when they do, they'll be in complete control; they'll know precisely what they're doing.

  I think: maybe they're the best people to explore the true Ensemble's mysteries, after all. Slowly, methodically, rigorously, respectfully . . .

  Po-kwai is successful on the second day; she seems pleased, but not surprised, by this. She's clearly gaining confidence in her skills with the mod, despite the obscurity of the operational details. How long before this growing sense of assurance, of control, invades her dreams - and shuts me out?