How would a Grey Goo robot compute given that it has no fixed form?

Question

This is something that confuses me with amorphous characters, monsters, robots etc... How do they compute when their particles don't have a fixed form? They are in a sense modular robots only taken to an extreme in size and number. Such a robot can freely transform, reassemble and other things. Take the T-1000 from Terminator Judgement Day. It can be blown to bits and always reform, that part is clear. What isn't clear is how it maintains memory and neural connections with such a makeup. Sure you can say that they do it wirelessly, but wouldn't there be too much interference and background noise?

This is somewhat addressed in the science newsjournal

"These soft circuit systems will act more like live cells, communicating with each other to form new circuits and moving around autonomously."

Basically recreating an organic brain, but if a Grey Goo robot were to be blown to bits it would lose its memory. If a Grey Goo robot were to have a computer ship inside it controlling it that could not be reformed. Many questions worth addressing.

If there was a way to make this more realistic by giving rules to the robots makeup it would be a big help, since I don't like handwaving everything into the story. The technological requirements as a advanced as necessary, since this is a very futuristic invention. And yes, the point of this robot is to reform if damaged, otherwise any other robot would do the trick.

Answer 1

Disclaimer

As you imply, the oft-depicted amorphous grey goo robot doesn't make all that much sense.

Humans and animals are made of nano-robots. But we still have bones, brains and livers. We are not a homogeneous mass.

The "grey-goo robot" is a hybrid of two (probably) scientifically incompatible concepts. The "grey goo" scenario, which is a semi-realistic idea that some kind of self-replicating sludge could grow out of control and start eating everything. (Note that in the most realistic versions the sludge can't actually do anything apart from spread like a algae, not much moving or anything). This idea is "bolted on" to a robot to make the typically imaged slime-monster.

Answer

But the slime-monster is cool. How to we justify it in a story that tries to play some kind of lip-service to scientific realism?

Several ideas occur:

1) The mind is located "off-site", and wirelessly puppets the body. In this case the best way of defeating the invincible monster is likely to interrupt or hack the signal. We will gloss over how it moves so well.

2) If a very high level of individual programming is assumed on each "cell" of the machine then the mind could just be distributed between them. Not at all like a human mind (where the neurons are made of cells), more like a computer network with tens/hundreds/thousands/millions of "neurons" in each cell. This requires some kind of futuristic sub-atomic tech (shrink-rays or something).

3) Very resilient cells, in a very floppy body. When the thing gets hit (or shot or whatever) the damage that the individual cells take is controlled not just by how hard the hit and how strong the cells, but also by how strongly they are held in place. Its the difference between punching through a piece of paper that is held in a frame, and one that is falling through the air. If the slime-bot was very "floppy" (liquid) then attacks against it would (at a large scale) appear to have quite a dramatic effect. However that same floppyness means that the microscopic components of the robot are likely barely being damaged at all.

Answer 2

Mixing the right amount of techno-babble and science, consider that in a neural network the output of the network depends on its layers, but do not change if you swap the neurons, which individually perform the same operations. What changes is the input/output of each neuron according to its position in the network.

Your "fixed form" equals the "fixed position", while being fluid your goo is capable of changing the position of the neurons while keeping the network architecture the same.

Answer 3

Modular computing, and flexible multipurpose components

First point to consider is that a computing task can be broken down into very small elements, and the physical location of those elements doesn't actually matter if their inputs and results can be routed to where they are going.

Think about how simple transistors make up an actual computer. Those are fixed in place, take inputs from one part, and their outputs link to another area. In modern computers those are all laid out on tightly packed chips. On older computers those could individual chips laid out on a board and connected with traces that you could see with your naked eye.

But what if those traces on a board were replaced with flexible wires? Then you could move and rearrange the transistors, to a limited degree at least, and the computer would continue to function exactly the same.

---

If we expand that into our grey-goo computer, consider if each component could reconfigure itself to meet whatever role it currently needs, or move itself to where other components need it to be.

Two 'cells' who need to communicate move apart?

• Throw an error and have a cell configured as a connector come in to fill the gap.
• Or, throw a different error, have one of the 'compute cells' move back closer to the other, can call for a different cell to take its old place.

---

Clusters of cells form 'compute-groups', tasked with solving some specific kind of data processing, and all compute-groups act as flexible controllers.

As long as enough cells remain as part of a compute-group that stores the core command and control concepts, then they can always issue directions to form any other compute-group that the colony requires to function.

So even if you blast away chunk of cells and disrupt local calculations and data, the rest of the pile will maintain enough computing networks and groups to recalculate, reform, and move on.

Answer 4

We already have self-forming microbot swarms in the laboratory. If you continue the micro-miniaturisation, or simply posit self-healing synapse systems, that should take care of your organic (or amorphous) greygoo.

Answer 5

It might be worth adding that when dealing with a technology level that assumes nano-scale replicators working in tandem, having any kind of a "microchip" as part of the body would be comparable to fueling a space craft with a wood stove. I don't think such a robot would have any identifiable control chip. If the nanobots that constitute this robot have enough self-contained energy and programming not to need external batteries or info storage, they likely do not need a discrete control module, either. All programming redundancy, energy sufficiency, and control mechanisms would fall to the greygoo itself. If, hiding in all that goo, there were boxy, large components swimming around and dodging shotgun blasts, the robot wouldn't be as amorphous or as resilient as it is. Might seem crazy, but you know what they say. Any technology, sufficiently advanced, is indistinguishable from magic.

Answer 6

The human brain already deals fairly well with adapting to losing surprisingly large chunks of itself.

It seems reasonable that a structure capable of re-aligning could do just fine assuming it didn't have anything else too complicated to think about while it was reforming the bits it needed for higher-order processing.

A housefly's brain is really small, but it's more than capable of responding appropriately to danger - You don't need much raw mass to provide enough intelligence to make decent decisions in clutch situations.

Even if the goo is completely disintegrated into tiny bits where no one bit is bigger than a houseflies brain, it can still eventually regain cognition by simply re-organizing it's structure. You'd inevitably lose data in this scenario, but surely any piece of grey goo large enough to do something interesting has a method of sharing information with the larger network wirelessly.

The real hand-wavy assumption bit comes in where you're talking about the manner in which the cognitive network maintains self-integrity during it's destruction and reconstruction - but it doesn't actually need to! Why would a pile of grey goo need to maintain a consistent linear concept of self-history? That sounds like a crutch. As long as, collectively, the entirety of all grey goo everywhere maintains some record of significant events it more or less agrees on within a certain margin of error, it can keep making rational decisions based on those events, regardless of the individual experiences of any cluster of cells that make up a "individual machine."

Answer 7

There are two major aspects: Computing and memory.

I'll forego the outer appearance, mostly, and just assume that the body consists of a swarm of cells that are capable of:

• Executing minor computations
• Storing small amounts of data
• Communicating over short distances (let's say a few cm), relaying messages. A larger group can group together to communicate over larger distances.
• Identifying by ID

Cells can be classified via ID. To prevent vast storage of IDs for each individual cell of a specific class, a pattern is used, depending on the amount of cells. For large amounts of cells, for example, "every cell with an even ID" would do it.

Computing

Computing is done in modular blocks. A small group of cells forms a processing block, multiple processing blocks can join together for more complex processing. Computing "foremen" cells take care of this: A foreman cell announces creation of a computing block, searches for cells closeby and creates a computing block with them while informing other cells about this action. Foremen cells would primarily save computer architecture.

Since the computing blocks can move freely in the mass of cells, disintegrate and reintegrate at will, communicate with other computing blocks, replace each other, be redundant and so on, it's easy for the swarm entity change shape, move through obstacles like prison bars, ignore bullets and so on - the computing blocks will replace each other or shift in place.

Memory

Every cell has its own memory unit, assume it's a few kb. In the whole swarm entity, memory is stored and accessed redundantly (by pattern) - basically a computing block can say "I require the data of a cell that is a multiple of X" and any of said cells can answer or relay the question. This enables information to be stored safely even if some cells get lost. Priority information is stored in more frequent patterns, less important situation in less frequent patterns - so critical information such as the system's OS won't get lost easily while random information MIGHT get lost when a significant amount of cells is damaged, but it is still unlikely. Like the processing blocks, storage blocks could be created as well to increase storage read/write speeds.

And... that's about it. You now have a cell swarm entity that can compute things and store things, so it can work and act arbitrarily well. It can also be split up, voluntarily or not, and reassemble. In emergency situations, where it is split up into many small groups and a vast amount of cells is unavailable, it can still grow together to smaller blocks, for a processing unit, call for the most redundant memory cells and start reassembling from there - it would be smaller, weaker, have less computing power and lose low-priority memory, but it would still work.

Answer 8

Each bot must be quite a bit smarter than we can make them today. Not nearly human-smart but, say, ant-smart. So, we have trillions of well-organized "ants".

The things the whole organism knows must be divided into small "facts" that can be stored in single ants. Each ant should be able to handle maybe a hundred facts and reason about them.

Important facts are stored in millions of ants, less important facts in just a few hundred. Losing some ants really doesn't matter.

At any given time, decision-making is located in a brain-like structure consisting of a number of ants close together. If these are scattered the organism will be confused and operating on very basic instincts until a new brain is organized. It may consist of any ants, not necessarily the same as earlier.

There is a many points here where we don't know how to do this yet, but nothing fundamentally impossible.

Answer 9

Computing is the EASIEST part of the Grey Goo Monster

We already have amorphous, distributed computing - it's called the internet, and it seems to work ok.

Computers don't need to know where other computers are in the physical world, they need to know where other computers are in the logical world. Generally, we use IP addresses to achieve this. There is a whole complex world of how we get messages from one IP address to another, involving various protocols and physical hardware.

Self Assembling Routers

Nanobots could talk to each other using the low level protocols, and if they detect that no router is present to connect them to the other bots, they spontaneously decide to make a router using the local bots, and then seek a wider connection.

The Real Challenges

For this to work, each grey goo bot needs to have an onboard processor, wireless or wired connectivity, and some sort of power source to move all those electrons around. ALL of those things are borderline impossible at the scale that grey goo implies.

Working with the Challenges

If I were working on a grey goo story, I'd probably look at what kinds of restrictions those challenges would place on my goo.

Processing power - If I isolate the nanobots from each other, do they become dumb enough that they can no longer function?

Power Source - Do the nanobots require constant exposure to some field (microwaves, X-rays, whatever) to power themselves, so putting them in a Faraday cage kills them?

There's lots of ways to work within the challenges to make interesting scenarios.

Answer 10

Maybe redundancy...the "memory" or state of the machine is represented in a distributed and redundant form, so in order to destroy it, you must destroy it completely, or at least more than a certain fraction (such fraction to be determined through plot points). That adds a certain twist to combat; just "aim for the head" doesn't work anymore, you need to go for a specific amount of carnage. (Maybe progressive? 80% damage destroys, but 50% damage just messes up some childhood memories?)