Is it theoretically possible to construct a self reproducing mechanical device?
Can robot or a device be made that, given the requisite resources, could construct and assemble a functional copy of itself?
To entertain this hypothesis further, let's add a set of explicit constraints or expectations:
- The copy must structurally match the original device as is realistically possible
- The device must be able to manufacture at least one copy of itself
- The overall replication process should sustain itself for an indefinite amount of iterations
- The device must not repurpose its own parts as parts for its copy
- The device must not peform any task other than replicating itself and preserving itself to be able to replicate
- The device may sustain damages of repairable or non-repairable nature as long as it obeys constraints 1 and 2
- The device may take as much time and resources as it needs to construct its copy as long as both remain finite
- The device may make use of essential external resources like electricity and cooling to sustain itself as long as its able to accommodate its copy to do the same
Can reproduction be emulated mechanically while obeying all the constraints above or is there a fundamental limitation stopping us from realising this concept? If so, what is it?
25 replies
If your definition of 'device' includes biological machines, that's us and basically everything else we call 'living.'
If you mean purely mechanically, as in with just cogs, levers, etc? Maybe hypothetically? But probably not really. If you ignore a lot of elements it might be conceivable to design a factory that uses mechanical computing to produce and manipulate stock materials being fed into it, but mechanical things are notoriously finicky when it comes to tolerances, so it'd probably take on too much entropy somewhere and wouldn't be able to repair itself.
If you allow electronic computers and actuators, it gets easier, but it's still extremely difficult to the point of practical impossibility to build any system capable of self-maintenance and self-calibration, let alone one that can fully produce every part of itself.
This post focuses on developing a purely mechanical device for the given purpose. Also it does not need to repair itself, it can sustain wear and tear as mentioned in constraint 6. It only needs to be able to create at least one copy that matches it structurally and functionally as stated in constraint 1 and 2.
Yes, theoretically of course it is.
Even practically it seems, they are all over the place.
They are not. Realistically this is not possible. And may never be.
Complexity is a real problem here. The more different things you want your machine to do, the more human intervention is necessary.
Let's give this idea the best chance at success and say it's raw materials and simple components are available to it. Basic fasteners, copper wire, sheet metal, etc.
This robot will undoubtedly have a motor, which would require a bunch of copper wire wound around a magnet, and assembled in a ring. This task is already done by robots. At least, the winding part is. The assembly is too, but it's a different robot. And let's say there's even a robot that can transport the wound magnets to the assembly robot. Such robots exist at least, this wouldn't be logistically efficient in this particular process but we're building a self replicating automaton here, sacrifices are necessary.
Anyway this one task requires at least 3 robots AND at present, the raw materials are being initially fed by a human. The magnets might come from a hopper but the copper wire is an issue. Let's say even THAT is surmountable.
The amount of floor space required for the process is probably something like 300 square feet. Just this process alone. Just to make a single type of electromagnet. We haven't built the housing, inserted the drive shaft, or any of the other required bits. just the electromagnet.
To make all the parts of a thing that a robot requires is going to be about the size of a factory. and that's assuming raw materials and basic bits are made somewhere else.
And even if you fully automate a factory, and it can make everything it needs to replicate itself, different machines would be needed to install them all somewhere.
You're thinking "but 3D printers!"
Stop. No. Just... No. are not that powerful. To be clear, a single metal 3D printer that's even half as capable as some people have been lead to believe they are could replace about 20 of the highest paid employees in my factory. They just don't work as well in industry as people seem to think. And that's printing in a single material. When you try to say, print a magnet with copper wire wrapped around it, you find very quickly that a: this isn't possible with the current technology, and even if it were, b, the durability just isn't there. It might be hard, it might be the right amount of brittle, but it doesn't withstand wear the way you'd expect for something with those properties.
The question was about theoretical possibility - practical limitations of present-day tech are irrelevant. OP also didn't put any constraints on how big, heavy or complex the robot would have to be, or if it has to be able to do anything else besides make more of itself. Even if the machinery it has to contain occupies a city block, that's okay.
It's a chemical and engineering problem. Plants and animals do this already. They are our examples to learn from.
A synthetic plant is a machine.
https://en.wikipedia.org/wiki/Self-replicating_spacecraft
Look up RepRap. A 3d printer that can print itself, except for the electronics
Except for the electronics, motors, hot-end, heater, and so on (motors might be printed partly I guess).
Indefinite can mean both not determined or infinite.
It can not make infinite of course, because at some point the Universe according to current standard model will cool out.
But apart from that no problem, as long as the physical conditions exist to support it.
Because you say mechanical that
doesmight preclude chemical reactions...At least a basic level of a real world self-replicating "device" in an abstract sense, you can "grow" some crystals given sufficient substrate liquid.
I think it is theoretically possible, and may be easier to realize at a microscopic scale since you can get a sufficient amount of liquid that can be considered a 'functionally-infinite' space and the device can more freely float around without having to worry about needing complex dextrous parts to put things in the right place.
A loose ball of dandelion seeds I put together would be my idea of a simple self-replicating 'device' that operates mechanically, in the natural world, at a human scale that I think meets your criteria. It floats around in the area and if there are dandelion fields where it goes, it can pick up the seeds, and if it bumps into a random obstacle in the landscape it will break into parts that can pick up other seeds. The only logical aspect of the device is how many seeds it is comprised of, and there is not a set sequence as to when it will break, because it just depends on where the wind pushes the 'device' into in the landscape.
Addendum: Condition 4 is hard to get around with my example, I can't guarantee that when it breaks apart only dandelions that weren't part of the original part would formulate the new part. Seems overly perscriptive if you count using its own constituent part to make a new part, then repairing itself up again as contravening that condition.
Precluding chemical reactions would pretty much make this impossible, given that any non-trivial machine would have to refine raw materials which would involve chemistry.
Permitting chemical reactions makes this trivially provable because evidence of self-reproducing chemical machinery is all around us.
You're absolutely correct on that. For the sake of the thought experiment and trying to keep to the spirit of the conditions, my dandelion ball device is internally, a purely mechanical device. The chemical and biological reactions related to growing dandelions and natural wind, all of that I consider external to the device.
They kind of already exist, see xenobots
Your link is missing the https:// so it doesn't work in at least some clients. Working link.
Oops, thanks, corrected!
Damn, that's nuts. It feels like a stone's throw away from programmable life
Isn't this a Von Neumann machine, essentially?
Learning this from XKCD was worth it both because it's an interesting concept and because it indirectly lead me to the Bobiverse.
Theoretically possible, and potentially catastrophic: https://en.wikipedia.org/wiki/Gray_goo
Constraint 5 seems like an odd addition, why wouldn't you want something like this to also be potentially useful? I would imagine if anyone does this for real it probably would break that constraint (as the xenobots do), because there's a good chance the builders would try to accomplish something with it.
This is solely to keep the premise of the simple and clear. The device only has one objective: to create a copy of itself. The problem only concerns itself with whether it is theoretically even possible to construct such a device.
I imagine with current or near future technology this would have to be a large automated factory with many different types of manufacturing elements within it. But I see no reason why it wouldn't be possible, just not cost effective yet.