"Permutation City" by Greg Egan - Review & Discussion

Egan is famous for his very high concept hard sci-fi fiction that leverages his academic background in mathematics plus deep knowledge of quantum physics and computer science. Permutation City is no exception, exploring an even broader scope of fascinating concepts than Schild's Ladder, which I'd read years ago.

It's one in a set of 3 novels, by him, grouped together as "subjective cosmology". But they're not linked in any way besides this categorisation of theme.

It pushes far beyond hand waving philosophy, weaving in some solid technical references, as it explores concepts of: continuity of identity, simulation of consciousness, extreme transhumanist mind modification, subjective reality, artificial life within deterministic systems, computability, parasitic computation and "dust theory". Many ideas I consider important to at least be vaguely aware of.

His characters are adequately believable and his writing style reasonably sophisticated. Although my attention/motivation struggles a little, these days, with this kind of classic novel structure: alternating chapters between several characters, telling a story largely in parallel.

Despite being published in 1994, 27 years ago(!), his 2045 and 2050 Earth settings still feel plausible. No glaring futurism details that failed to transpire. There is kind of an absence of smart/mobile phones, but not really.

The use of centralised computer servers, for (extreme) processing power, might echo back to a time before the personal computer boom. But it actually fits fine with contemporary cloud computing, in 2021; our consumer electronics increasingly pushing towards Egan's "terminals". Even for the highly response dependant application of game rendering, with Google's Stadia and Amazon Games (which are admittedly struggling, but for other reasons, perhaps).

There was one particular instance, that bothered me, were Egan knowingly contradicted a certainty of computer science. He explains, in an FAQ on his website, that this is for the sake of a simpler (more approachable) narrative. But it's not possible to discuss this, or other details, without some big plot spoilers...

So in terms of basic review, I'll just say that this is a recommendation from me - it lived up to it's reputation.

► Specific Discussions [BIG SPOILERS!]:

• Parts - The book is split into halves. I didn't even want to allude to this, above, because merely knowing the setting of the second part means that Paul Durham was correct and successful in his (highly questionable!) aims, during part 1. 

They succeed in sending scans of their brains, along with a conscious digital instance of his mind, to a simulation of reality running on a "TVC" (Turing, von Neuman, Chiang  - a fictional computer scientist) cellular automata universe, that computes itself into existence, somewhere in the "dust". Forming an ever expanding substrate to run a simulation of a communal city, various private domains, and an entire planet with evolving artificial life using fully deterministic physics.

• Digital hell - One of the private sections is apparently dedicated to personal torment, for a guilt ridden German billionaire. I found his backstory and self torture a little tiresome, to be honest. But simulated suffering is a potentially cosmological scale moral hazard that's increasingly important to think about, going forwards. 

It's likely that bad actors will, at some point, command the computing resources to torture digital consciousness (re)creations in complete secret. Maybe even incidentally, or accidentally, once computing reaches unimaginable scale, in the mid-future. Even without something a dreadful as Roko's Basilisk being realised. 

• 2045 setting: Perhaps coincidence, but was he inspired by Kurzweil's futurism predictions for the Technological Singularity occurring around this date? Back was that timeline the same in the end of his "Age of Intelligent Machines", as with the 2045 date featured of his 2005 tome "The Singularity is Near)?

• Slowdown factor: It's interesting that Egan has his 2045 simulated minds all running at a rate significantly *slower* than real-world time. I think sci-fi typically goes for far faster, as it's more exciting. 

Although, it seems silly that he's so specific about it being 17 times slower. A fundamental limit, even for the super-rich. But not specifically attributed to the types of calculations or hardware limitations, because it's said not to be improving over time, with Moore's law (like) doubling in performance. Apparently the total customer demand for computing power keeps exceeding the supply. 

This aspect feels ominously relevant to today, where a goldrush of ever-more crypto-currency mining has gobbled up the supply of desktop GPUs (plus malware hijacking of machines, etc). Keeping the prices high. Price performance gains have been modest in other computing parts, too, like spinning hard drives for years.

More interesting, the upload dystopia of those without huge wealth, who have to run their digital minds at an even more substantially slower rate. Their reality paused without warning, when computing resources are bought out from under them by big-money. Though not harmed, they skip forwards in time. And the combination of the two (mostly the slowdown) make it impossible to connect to those in the physical world. Who don't really want to make time for messaging their dead, let alone visiting. Ghettoised by time dilation.

• Incomputable (scientifically impossible!): the experiments Durham runs on a simulation of himself drag on too much; a lot of repetition of counting to 10, etc. In the last of these "tests", his mind state is supposedly computed, on many different servers all around the world, out of time sequence order! This was the specific problem I mention in the review, that really irked me.

Given that a mind state evolution is almost certainly NP-complete (like turbulence flow in fluid dynamics, etc), there's no way a system could just spontaneously arrive at an end state (and then go back and do the calculations in between). Even with deterministic systems, like simple cellular automata (e.g. Conway's Game of Life), the end state can't necessarily be found without calculating every step in between, from the start, forwards. 

But this is Durham's supposed inspiration to arrive at....

• "Dust Theory": that any self-aware pattern of information (e.g. a human mind, or simulation thereof) will always find itself. Even if that pattern is only spread out through the correlation of states of the atoms of dust spread throughout the universe, at different times, etc. 

Somewhat like the anthropic principle - that we see a universe with physics constants (electric charge, gravity, speed of light, etc) exactly as they are, because if they were too different, planetary formation, organic chemist and evolution would not be possible, so there would be no observers to see this.

Again, in literal terms, dust theory makes no intuitive sense, to me. To simply cut the causative web away and have no evolution between consecutive states, just the states themselves, only seen by looking in an implausibly dispersed manner. Like finding the works of Shakespeare by plucking non-adjacent letters from a library of other books. It's meaningless.

Thinking of the quantum branching of (parallel) universes might make more sense. Like, Durham kills himself, so that his consciousness wakes up in another world, where his memories are miraculously consistent. But it's more likely that he would perceive events that *stop* him committing suicide in the first place. All it takes is the possibility of one random fluke, in the gestalt of all possible universes with him trying to kill himself, because all the dead versions don't perceive anything at all.

I think that an Omega Point Singularity (or equivalent) is a more sane mechanism for actual (digital) reincarnation. Although the book's plot in part 2 would look the same to the observers, in either case.

A personal aside: decades ago, I found a similar micro-scale anthropic principle thought experiment: could one change the speed of light in the universe? By having your mind simulated on a machine that utilises the speed of light, requiring it's speed remain true for it to continue working. Then retuning the machine's mechanisms, such that the mind simulation could only continue to keep existing in a reality where the speed of light decreases/increases. Obviously only perceived by the simulated mind.

By extension, I wondered if similar could be achieved by playing the lottery, with a suicide machine only permitting you to live if you won. There'd be a *lot* of universes where everyone sees a dead idiot, but the idiot would only exist in the instance that they won, so they'd be guaranteed to see that. If the mechanism were perfect. And intervention was not more likely than a lottery win.

I certainly didn't have the inclination (or equipment) to try to find that out! But at the time, alone depressed and fairly desperate, at the end of my failed physics degree, with failing cognitive ability due to encroaching ME/CFS, no substantial medical help forthcoming, I wondered if I might need a lottery win to stand any chance of finding a way to fix myself, and continue to exist in a meaningful way. Or, mentally contorting even further: maybe, if one resolved (before hand) to dedicate most of the proceeds to a specific cause. One for which its success would be necessary to avert the future end of humanity. Then maybe that would precipitate a win...? 

Well, I tried picking random numbers, online, with a home made program that fully hid the numbers from me until after the draw (before there was a lucky dip feature). A kind of Schrodinger's lottery ticket. Although, that would require quantum coherence over macro (massive) distance scales. Obviously I didn't win. Lol!

Well, the vast majority of version of me didn't. If many worlds interpretation of quantum mechanic is real. Funny to think of there being versions of yourself that did win big, out there in the multi-verse. But my personal feeling is that reality is more deterministic than that. Quantum mechanics not truly random, etc, once we fully understand it from abetter perspective.

• Artificial Life: was the buzzword of something I was kinda hyped about around the start of the noughties, with routes back to the 90s. Genetic algorithms in 2D tadpole-life agent simulations, simple cellular automata, and later Conway's game of life. 

The character, Maria, is an artificial life tinkering addict, who flukes finding an years elusive means to have a simple bacteria equivalent organism evolve the ability metabolise a different pseudo-chemical substrate, in their fictional "Autoverse". A conception of a physics that is fully deterministic, and arising naturally from a specific cellular automata rule. So it's massively more efficient to compute than the equations of real physics, but consequently with a completely different (though somewhat comparable) set of chemical elements arising from the emergent physical constants, etc. 

I thought, originally, that her fluke evolution was going to be down to her simulation being suspended mid way through (when all servers are bought out for a test of ). Linking to Durham's philosophy, somehow. As someone else reports being unable to reproduce her published work. But that doesn't seem to be the case.

The conscious lifeforms that eventually evolve, in part 2, on the simulated "planet Lambert" are kind of boring and unimaginative. Almost literally the joke physicist's approximation of a spherical cow! Extremely simple body, but apparently super-efficiently wired nervous systems/brains that allows individuals to be the size of a fly. Conscious as swarms, communicating concepts in dance. And somehow magically arriving, and confirming scientific theories between members of their society, without ever developing technologies to probe their reality directly. Which really is silly; the lack of feedback.

• Falling away: the way in which their Permutation city melts away is kinda of cinema theatrical. Although hard to say how it would go. With the artificial lifeforms changing the laws of the universe out from under them, hmm... It almost ridicules itself. Conceptually. Like, whoever has the strongest dependency on impossible changes to reality will crush the other's existence. Surely they'd each perceive themselves to win-out.

• Monolithic imagination: the lack of variety of human existence in the TVC city is a little disappointing. We only really see the city that was designed and created by human contractors, back in 2050. Big buildings, but conventional looking. Weirdly composed (digital) post-human descendants are described attending the meeting about the fate of their universe, but of course it's too difficult to explore the potentially very diverse realities they might inhabit. This is a perpetual frustration with all hard sci-fi - the fundamental undecidability of the future makes it literally impossible to even predict meaningful what type of things are going to be possible, out beyond a few decades.

• Self-modification: I liked that we explored a fairly extreme philosophy of digital human existence. The "solipsist nation" lovers. The guy experimenting with radically rewiring his (digital) brain, to simply choose to be happy with situations. (The kind of thing that entirely didn't happen, a missed opportunity in the Bobiverse trilogy, that I recently criticised.)

To go beyond that, as almost an art form, to make oneself eternally content within a single simulated moment. By carefully tuning the parameters of his working and long term memories, his digital body tireless, etc. If it repeats perfectly, with no memory of the repetition, do you even exist for more than a single moment, objectively? You'd need an external interrupt to be released from such halting state.

To be clear, I find the idea of such an existence ugly and perverse in it's wastefulness. To avoid ever *quite* halting forever in a loop, by contriving new pointless things to be happy doing. Compared to an aim of perpetual change and growth in complexity of thought and function. Obviously impossible to imagine what kind of things that might ultimately involve, though. And technically, if you do have literally forever, one would *eventually* have to grow as a person, provided you kept changing for eternity.