Why is time-dilated simulspace not an I Win button?

Question

First, a bunch of relevant quotes to set the stage.

What is simulspace?

Virtual reality overrides the user's physical senses and places them inside an entirely computer-generated environment called a simulspace. (EP 240)

What is time dilation in simulspace?

Time itself is an adjustable constant in VR, though deviation from true time has its limits. So far, transhuman designers have achieved time dilation up to 60 times faster or slower than real time. [...] Time slowdown is more commonly used, granting more time for simulspace reacreational activities [...], learning, or work (economically effective). (EP 241)

Sounds great! Does it take a lot of effort to get this going?

Dedicated networks with high-capacity information processing are required to render and run large and complex hyper-real simulspaces with many users [...]. Smaller simulspaces capable of hosting a smaller amount of users can be run on a smaller distributed network of linked devices.

Is it hard to get into VR? Well, ideally you'd want a special pod that'd prevent you from flailing about and hurting yourself, but basically...

Characters can enter and leave the simulspace at will, but toggling in or out takes a Complex Action. (EP 262)

How easy is it to communicate with the outside world from a simulspace?

A character accessing a simulspace may still interact with the mesh (and through it, the outside world) assuming the domain rules allow for it. Any outside interactions are subject to time dilation issues, however. For example, in a simulspace running faster than real time, holding a chat with someone in outside meatspace is excruciatingly slow, as real-world seconds translate into minutes in VR.

Now that that's established

And that's largely it for rules regarding simulspaces. They are mentioned in the chapter on psychosurgery, but not really anywhere else. Now, from these quotes (and the rest of the text) it appears that one could make themselves a barebones personal office simulspace that'd be easy to run either on just their mesh inserts or at most a small extra device. From the way it is described, it also seems that time dilation by the factor of 60, while an upper limit, is not hard to reach.

One could drop into this simulspace at-will, spending a grand total of 3 seconds on this act, at which point they'd get the benefit of a minute of thinking to every second in meatspace. Granted, they can't effectively act in said meatspace in that time. However, that seems like barely worth mentioning for the benefit it provides. Cost: 0(-ish), mental actions gained: 59

Now, compare it to a few similar things. Informorphs have speed of 3 (2 extra actions per action phase; cost: no physical body). Multi-tasking implant runs short-lived forks that can perform 2 extra purely mental actions per action phase; cost: High. Mental speed nanoware's description is actually similar to the effects of time-dilated simulspace, and grants 2 extra actions and a bonus to Initiative; cost: High.

Time-dilated simulspace clearly wins. An argument can be made that it won't be quite as efficient, because outside limitations such as processing time of hacking software (not that you would use such a thing, of course) would become an issue. This would be left very much up to GM in every given situation, as is the habit of the system. However, that's clearly an issue other mental acceleration methods would love to have.

This cheap trick doesn't appear to be "official". Not only is it wildly out of line with similar effects, it's not even discussed as a possible approach in Transhuman's chapter offering advice on combat hacking.

The actual question

My current interpretation of the rules is clearly untenable. What factors limit the utility of time dilation, making it less of a magic I Win button? If the existing material doesn't explain this, what possible modification to such simulspace can be made, that preserves its use for accelerated psychosurgery and skill learning?

Answer 1

To begin with, I'll point out one important assumption we can safely make.

Processor power is essentially a constant for any given computer.

A system which can run one person at the maximum 60x dilation could, equally, run 60 people at normal rate.

An obvious and important corollary: any system that cannot run multiple transhumans in a simulspace is also not powerful enough to run an accelerated simulspace.

This is the key understanding you need, I think. Your second and third quotes, put together, imply the problem you raise - but there's no justification for assuming that they go together that easily. Indeed, quite the opposite (p242):

On rare occasions, office environments are run in simulspace with time dilation...

...Since this requires the workers to access a centralised wired network and leave their bodies unattended ... (it) is typical only of some governmental installations and corporate habitats.

Nowhere does the book imply that a small mesh-distributed simulspace can reach the maximum time acceleration, or anything remotely close to it. A "centralised wired network" is required. (Regardless of processor power, wireless mesh communication probably doesn't have the bandwidth required for that much accelerated input and output.)

Now we can assert some limitations on the effectiveness of this technique in roleplay:

• It is reasonable to rule that time dilation is impossible - or limited to a very small factor - on a simple mesh insert. The typical mesh insert is barely capable of accepting one ghostrider; it doesn't seem likely that it has enough spare processor power to significantly accelerate your mind. (Note that a certain amount of accelerated computing is already implied in uploaded minds; infomorphs get extra actions. If a typical local device could handle more than that, infomorphs would already do more than that.)

• The 60x accelerated network requires a dedicated, physically wired system, and the EP equivalent of serious server hardware.

  Of course, the EP equivalent of serious server hardware is still small and might even be made portable, but it's still an investment. VR entertainment sites probably do something like this... but then, they have to invest in pods and support hardware anyway; the computing power is trivial by comparison.

• Large time acceleration requires that your body be protected by a pod, not just left on the couch. (Spare nerve impulses leaking out at 60x acceleration probably looks something like a really serious epileptic fit, and then some - unless specialised hardware protects you.)

• Optional, but plausible: you could rule that the maximum 60x acceleration (or anything over, say, 10x) is only possible when fully uploaded or using a synth brain. (People in flesh morphs are still using neurons to think with, and even with nanotech assistance in a menton or remade morph, those only run so fast. The human brain probably can't just be accelerated beyond a lower cutoff while it's still embodied.)

The most important limitation, however, is this one:

• Hyper-accelerated thinking doesn't help much in most roleplay situations.

  It's not an "I win" button unless it actually, you know, helps you win.

So let's look at some Eclipse Phase roleplay situations and see what happens:

• Combat. Dropping out of your body in a firefight is an extremely stupid thing to do, so a transhuman that tries this will be taking no extra actions, since their morph will get shot while it lies there. If the hardware's available, an infomorph party member could sacrifice a round to take 60 actions the next round... but what's she going to do with them? She can't take most actions that would meaningfully effect the fight. (She could send a lot of email requests for help, though.)

• Diplomacy. In social situations there's no gain, since you have to run at the speed of whoever you're talking to.

  You could have an accelerated party member gathering information and mailing you hints as you talk, but that's not likely to be gamebreaking - just a fun extra tactic.

• Research, or other brain work like office work. Well, you can certainly get a lot more done that way. But again, it's unlikely to be gamebreaking. Remember that most corps already do this; it's part of their commercial advantage.

  Since most office work or research involves collaboration, to make meaningful progress you need to accelerate everyone involved. Which cuts them off from social interaction with anyone else during the project. (This suits corps just fine, of course.)

  You can do this privately... but this is Eclipse Phase. Everything is data. So consider the consequences. If you set up dedicated high-speed hardware and start working on it... a bunch of indentured infomorph data miners notice. And somewhere, a corp industrial spy starts wondering what you're doing that's so interesting. And whether his bosses would like to have it.

• Hacking. This example does need addressing, because by a strict reading of the rules you could take a lot of mental actions very quickly to speed-combat-hack everything in sight.

  But I'd argue against this interpretation, because a hacking action is an intrusion - subtly, carefully, exploiting weaknesses. If you suddenly throw 60x as many attacks at high speed... well, there's already a rule for that. (See "Brute-Force Hacking" on p256.)

  Time acceleration would let you plan a standard intrusion much more quickly. But I simply don't think this is all that game-breaking: it won't let you do anything you couldn't already; it'll just let you do it in an hour instead of a day or two.

  You'd be well within reason to impose a penalty - anywhere from -10 to -30 depending on the simulspace - to all of the Infosec tests as well. Hacking involves dealing with the raw data and exploiting weaknesses. But a simulspace is a drag on this, since all your mesh interaction gets filtered through the simulspace's UI, instead of your usual tools. You're putting an extra interaction layer between you and the metal. Hackers hate having to do that.

  (You could probably design a custom hacking simulspace to reduce this penalty.)

There are also social factors to consider:

• Relying on acceleration for social purposes is pretty gauche. It's probably considered even more unimpressive than a person who keeps looking away to look up references while speaking.

• There is a processing power shortage, caused by the infugee crisis.

  Millions of people are still in dead storage - archived but not running. (This is one reason for corp exploitation of the infomorph labour pool - if you're not earning your processing power, they can drop you back to inactive and revive someone who will.)

  That means millions of people are, literally, dying for a processor to use.

  In some societies (I'd expect Mars and LLA particularly), using up that much simulspace time to accelerate yourself will be resented by transhumans, especially infomorphs, who can't get the same resources to reactivate friends and relatives. They'd rather revive 60 friends than run you at 60x speed.

Answer 2

Essentially what is being suggested is sixty egos/forks operating simulatenously.

I would like to incorporate some of the insights from shared memory and distributed memory parallel programming (OpenMP, MPI respectively) and high performance computing into this.

A related topic is the following:

Cybernetic brains are where the ego (or controlling AI) resides in synthmorphs and pods. Modeled on biological brains, cyberbrains have a holistic architecture and serve as the command node and central processing point for sensory input and decision-making. Only one ego or AI may “inhabit” a cyberbrain at a time; to accommodate extras, mesh inserts (p. 300) or a ghostrider module (p. 307) must be used.

Essentially a Ghostrider rmodule is an habitation for a Cyberbrain.

To treat it roughly, it would take at least sixty of these to provide the equivalent of a 60x speedup (a slowdown, of course, is a lot easier, just put in sleep statements in your code). Basically, by using the same processing power as a cyberbrain/module you could achieve one 2x speedup if, and only if, they are engaging in entirely independent actions (fork-join OpenMP model).

However if there is necessary communication between the two paths (communications world, MPI model), then (simplifying) every two egos in communication will require an additional equivalent computation power (mainly high speed interconnect, processing) plus additional programming.

Rather than 60 an upper known limit of 64 is recommended, just because :)