Quicksilver

TheMite: “I had the misfortune to try one of the cheap Starware Zand models recently. It was pretty slick on the lunar surface, and the feeling of warm regolith flowing past underneath was one of the more enjoyable somatics I have experienced recently (see attached XPmark). The very low profile was a nice bonus: I could sneak up on a sentrybot very closely before it noticed me visually. It was then I discovered that in low gravity the droplet homing is next to useless: sure, the droplets will get back to you, but they fly far away and since they have to roll back they won’t be available for a significant while. Worse, this model had no hidden mode on the homing and the sentrybot started to target them by their navigation queries. I ended up scattered over a big area before I could withdraw, mostly by sheer luck.”

ICS Knology: “Sure, the Zand IV and Prism are pretty lousy if you want to fight. Or if you are a singlethreaded intelligence. For low gravity use we would recommend an Exotech Quicksilver Dea instead; you can get ion drive units included for a small added fee, and they have some nifty waypoint homing macros that cut down signals emissions. But as we said, this kind of morph isn’t the best if your ego is originally single-threaded.”

Teodor: “Xigki are apparently working on a heavy duty Quicksilver, with stronger unit interlinks that would be able to reach human-level or beyond strength. Likely a bit slower but more resilient thanks to all the diamond walls and locks. According to industry rumour they are trying to compete with Starware in the industrial flexible synthmorph market.”

Hmm, so what are the munchkin problems with my design? Just the EMP resistance? I guess that depends on how deadly you want to make EMP for swarm systems.

In game, I can see plenty of uses for a granular material morph in terms of manipulation flexibility, speed of movement and especially power distribution: having all swarm units run their own power units misses the economies of scale of having a flexible internal energy distribution system. Swarmanoids would seem to have a rather short range before they need to replenish their internal resources.

(I love munchkins. I usually annihilate them utterly in my games by giving them what they want... )

Interesting. So in what way is a swarmanoid humantech but a liquid bot TITAN-like?

A true "smart liquid" sounds hard to make with transhuman tech. Although one might make something like an amoeba, filled with a cytoskeleton of microtubuli/actin fibers (or a nanotech equivalent) that could crawl using pseudopodia. Put control into embedded micromachines that form an amorphous computing network, and you could at the very least get an interesting kind of liquid/gel bot. Doing high performance bulk computation (like running AGI) is likely hard for transhuman tech, but motile and sensing blobs is probably on the level of a school project.

Amorphous computing is by the way quite relevant to EP, since this is likely how many nanosystems are organised. See
http://en.wikipedia.org/wiki/Amorphous_computing
http://groups.csail.mit.edu/mac/projects/amorphous/
http://isandtcolloq.gsfc.nasa.gov/fall2002/presentations/abelson.ppt
Lots of little local simple processors sending short-range messages, forming patterns of activity or making products with large-scale order.

I don't mind adding game balancing tweaks to the morph. But I prefer to make them so they make physical sense.

But if you are an undulating flat sheet, you should be able to swim very well. And if you anticipate the hydrodynamics you should be able to swim faster (however, that also suggests that the normal swim skill might be useless and one better use skills such as Swim: Quicksilver or Physics: Applied fluid dynamics).

Maybe the shapeshifting abilities were not clear enough in my writeup; I think many people assume it is a proper liquid rather than a granular conglomerate. What I have in mind is very much like the system discussed in
http://www.niac.usra.edu/files/studies/final_report/883Toth-Fejel.pdf
Note that systems like this do not make use of antennas to send signals from unit to unit, but actually have small optical connectors between neighbouring units. Antennas are still there since detached droplets need to be guided and everything wants to be on the mesh, but they are not absolutely necessary for function. This is why I think it would be pretty EMP resistant.

The argument that EMP is particularly nasty against nanomachines also do not apply since we are talking macroscopic (mesoscopic?) devices here: if EMP is deadly to devices like this, then we should also assume flexbots and anything with a bush robot hand to be broken by EMP. Which actually makes sense: I think the claim in the core book that since most devices use optoelectronics EMP has few effects on normal technological environments might be wrong. Advanced technology contains extreme numbers of tiny components (not just antennas) that could be affected. Nanocilia used to transport things inside the self-repair system, the long conductive paths of a chameleon suit, the metamaterial shielding (*especially* metamaterials, since they are essentially a crystal of antennas!) - all sorts of devices depend on potentially vulnerable subsystems.

Maybe. But there is no real speed advantage here, since optoelectronic signals and radio are nearly equal in speed. Radio also has bandwidth limitations that are quite severe compared to parallel bundles of optics - not to mention the amount of easily detectable radio noise a constantly updating distributed mind would produce.

In the case of this morph (and the swarm) another limitation by the way is that the odd shape and ways of locomotion are very alien to most transhumans, making the integration test much harder. (Exotic morph, -30).

In order to work as a body the units will have to use "macros" to do things: the controlling ego doesn't give detailed commands where every single unit is going, but rather general goals like "I want to go that way", "push towards that", "undulate at 3 Hz with a wavelength of 30 cm and amplitude of 4 cm", "extend a pseudopod through that hole". If you had to give detail commands nothing would work (just like a human couldn't control their body by giving direct muscle group twitch commands; the motor cortex actually turns the prefrontal goals and macros into motion sequences, which are further refined as you go down the spinal cord). These commands then get interpreted by the morph software and turned into working "games" (moves of units relative to each other, either just flexing of connectors or actual detach/move/attach sequences that changes the shape). There might be hooks for running your own macros and detail commands, but that is for the experts and morphnerds - most users will want to move and act, and that's it.

So I doubt the ego will experience its body as lots of units as long as everything is fine. That is the whole point, otherwise it would be a swarmanoid. Of course, when things go wrong the complexity of the body becomes far too noticeable...

("Help! I have splashed and I cannot get together!")

If we assume cubes, each will at most have six connections. A 0.1 cubic meter Quicksilver will have about 3.7 million units if they are 3 mm sized. So a message from one side to another from a puddle-shaped form would pass through about 666 units, which might be a bit much for current internet routing protocols but doesn't seem that hard to implement with a more specialized protocol. Since distance is closely linked to physical coordinates you can actually make routing relatively efficient. Still, if every unit wants to send a one-bit message per second to every other unit that is 3.7 megabit/s per unit, and each unit will on average be between at least a few thousand units (some geometry complications here) - so we should expect a gigabit/s through each unit. Doesn't sound impossible to do with EP tech or even near future tech (with wavelength-division multiplexing we can currently do terabits per second over optical fibres), but it is a hassle.

A more plausible calculation would be based on the human brain. Assume we distribute the 10^11 neurons across the units, putting about 27,000 simulated neurons per unit (about the complexity of a dumb insect). That is a few cortical columns and might actually be a pretty convenient simulation size (lots of local connectivity gets done inside the unit). About 10% of the connections are long-range and would link to other units. These need an update rate of at least 1000 Hz to think at human speed. So that means each unit will send and receive 27 gigabit per second. Sounds like I might need those terabits per second after all (Or, more sensibly, clump neural processing into specialized units, have a few others do morph control and have most nodes just obey the brain units)

The basic design challenge is to keep long-range communications rare enough. This isn't unsolvable, I know some tricks from neural simulations on supercomputers (which are not too different from the above scenario: you want to lump computations cleverly on processors to minimize communications) that can give you orders of magnitude increase (e.g. just send action potentials and not "nothing happens" messages). But I guess it is overkill for a roleplaying game to try to optimize the implementation... I'll do that when I try to *build* one of these beasties sometime near the singularity

I think radio has one advantage: it can broadcast things - at a low bandwidth, but with no need for routing. I would probably use this mainly for homing between disconnected units and *especially* for the handshake protocol merging droplets. You don't want to accidentally merge with the enemy... (And here we get a beautiful argument for why EMP still is a major problem for the Quicksilver - as long as it doesn't separate it will be fine, but having been EMPed it will have a very hard time rejoining parts until it is repaired. It will essentially lose its "regeneration" after one or two EMPs.)