Self replicating (with remote human supervision) robotics for base on the moon is merely a very big engineering project. I’d say somewhere on the scale of modern operating system with the associated software (word processor, web browser, etc). Perhaps that plus the hardware and fab plant design, on the high estimate side.
But we have plenty of unused land and sea floor on Earth; the least hospitable place on Earth (excluding active volcanoes perhaps) is much easier to live in, than the moon.
I don’t think so. Software is much easier than hardware. In fact the “self-replicating robots” problem probably contains a specialist operating system subproblem in it! One thing I learned watching collaborations between mechanical and electrical engineering PhD students, is that mechanical engineering goals are generally much more humble.
Mechanical E: build a good helirobot engine, Electrical E: build a fancy helirobot controller. (Building a new OS is also a PhD thesis in software, e.g. of roughly comparable difficulty).
Exactly. Self-Replicating robotics on Earth is a global instant victory condition. Completion of one would result in machines that could double their production exponentially, leading to practically infinite production capability within no time.
Suggesting self-replicating robotics is akin to saying we should just solve this whole not being post-scarcity problem.
Self-Replicating robotics on Earth is a global instant victory condition. Completion of one would result in machines that could double their production exponentially, leading to practically infinite production capability within no time.
This does not follow. This depends on a lot of conditions, such as how fast the robots replicate, what resources they need, and how broad the circumstances they can replicate are. If for example someone could make a self-replicating robot but the robot required boron in some critical part, its replication would be severely hampered.
Moreover, even having a self-replicating robot isn’t by itself necessarily useful if you can’t control it in detail or get them to then do exactly what you want. And a self-replicating robot with no control could be quite bad.
But these are essentially minor nitpicks, and I agree with your point if one adds the appropriate minor disclaimers.
I was thinking more in terms of the original claim. Self replicating robots able to replicate quickly enough and flexible/controllable enough to make a permanent colony on the moon for us.
I mean, I assume that was the original point instead of sending very large, slow Von Neumann machines to tile the moon with copies of themselves. That would be cool but probably not worth the expense, and it’d carry such an awful risk of backfiring on us.
Exactly. Self-Replicating robotics on Earth is a global instant victory condition. Completion of one would result in machines that could double their production exponentially, leading to practically infinite production capability within no time.
Per Robin Hanson, a machine shop can put out its own mass in equipment in roughly a month or two. And yet, the economy doesn’t double every month, or even every year. Why not?
There seems to be a fair chance the reasons are mostly rooted in cognitive biases, cumulative coordination mistakes, economic rent-seeking, and so on—not anything technological.
A well planned lunar or orbital mission might well be free of these issues. Space conditions are mechanically simpler in some respects, so there’s a stronger case for pre-planning everything rather than requiring a market economy to make it work. Supporting structures are less needed, transit is less two dimensional, and solar energy can be harvested at scale with low costs in equipment density. There is also instant access to ultra-high vacuum conditions which are useful for refining. And in addition to the endless cheap sunlight, there’s no anti-nuclear lobby which can claim it’s in their back yard.
Suggesting self-replicating robotics is akin to saying we should just solve this whole not being post-scarcity problem.
Maybe we should solve this whole not being post-scarcity problem...
Per Robin Hanson, a machine shop can put out its own mass in equipment in roughly a month or two. And yet, the economy doesn’t double every month, or even every year. Why not?
If we switch the example to an excavator which outputs its own mass in an hour or two, does the answer to your question become clearer?
A quick process like that is pretty much insignificant compared to a month or two, let alone 15 years. Unless there are tens of thousands of other steps in the chain of comparable length, it doesn’t come close to explaining it.
As I see it, there are roughly four steps:
Excavating.
Refining.
Power collecting.
Manufacturing.
The ones towards the end seem to be the biggest time sinks. However, power collection should not raise it by more than a factor of two or so. I don’t think it takes many months to mine enough coal to pay for the energy costs of coal mining equipment, for example.
I think you underestimate how much work did go into e.g. Microsoft Windows, or Linux (complete with drivers, graphical environment etc etc); I meant that by operating system (not PhD project sized of course. There’s giant difference; you can call both ‘operating systems’ just as you can have mechanical engineering project of making ‘a liquid fuel rocket engine’ and it’ll be so much smaller than Space Shuttle main engine). Also I included the design of computer hardware and fab plants on the high range.
The engineering of the self replicating factory, with the exception of engineering of few components, is mostly down to building system out of components, which is more like software engineering and less like design of a space shuttle engine (or for simpler analogy a ball bearing).
It’s untouched ground, there’s a minimum threshold for being a self replicator, but no minimum threshold for getting called ‘operating system’ (I myself can probably write an ‘operating system’ in a week or two, or even in a weekend, depending to how little we consider to be an operating system), and such replicator—I am speaking of supervised self replicator, i.e. largely under remote control—is economically ineffective compared to regular, staffed factory. A replicator is no magic—it is a moderately big complex of factories, with automatization of even the rare tasks that would normally be much much cheaper to perform with humans—there is literally no commercial incentive to design something like that, considering that it would cost billions to design. More, actually, because the hardware of the first original is going to be awfully expensive.
I must make clear what replicator I am speaking of. A practical one, which is to significant extent remotely controlled, and for which much of the difficult to manufacture electronic components (CPUs, memory chips) are shipped from Earth (they are lightweight). Such replicator would require multiple extremely heavy launches to get there, it would be very fragile, and it would grow considerably slower than even remotely wise investments. So you’ll probably be better off simply doing something else and going to moon later.
It is not really glamorous awesomeness by any means. It would largely rely on existing CNC machines and existing robotics, it would be huge, it would be messy, and it is going to break down once in a while at first (with a bit of luck that may be resolvable with remotely controlled robot, in a manner similar to trying to solder chips onto a board by holding them with chopsticks).
edit: tl;dr; it can be done, it wouldn’t be the most complex task mankind ever done, and it is going to suck.
I am calling you out—I don’t think you can write an operating system from scratch in less than a few years. (I thought tinkering in an existing toy OS for a semester in my undergrad OS class was fairly hard. I once tried writing a “real” file system from scratch—very difficult also).
How familiar are you with software engineering or mechanical engineering? You are systematically underestimating the difficulty of engineering tasks, I think. You sound like a futurist, not an engineer.
I am not sure I understand the concrete proposal you are making, nor am I sure how much “replication” should be a part of such a proposal. A lot of people are thinking very hard about economical moon/mars colonization, any thoughts on their proposals?
Self replicating (with remote human supervision) robotics for base on the moon is merely a very big engineering project. I’d say somewhere on the scale of modern operating system with the associated software (word processor, web browser, etc). Perhaps that plus the hardware and fab plant design, on the high estimate side.
But we have plenty of unused land and sea floor on Earth; the least hospitable place on Earth (excluding active volcanoes perhaps) is much easier to live in, than the moon.
I don’t think so. Software is much easier than hardware. In fact the “self-replicating robots” problem probably contains a specialist operating system subproblem in it! One thing I learned watching collaborations between mechanical and electrical engineering PhD students, is that mechanical engineering goals are generally much more humble.
Mechanical E: build a good helirobot engine, Electrical E: build a fancy helirobot controller. (Building a new OS is also a PhD thesis in software, e.g. of roughly comparable difficulty).
Exactly. Self-Replicating robotics on Earth is a global instant victory condition. Completion of one would result in machines that could double their production exponentially, leading to practically infinite production capability within no time.
Suggesting self-replicating robotics is akin to saying we should just solve this whole not being post-scarcity problem.
This does not follow. This depends on a lot of conditions, such as how fast the robots replicate, what resources they need, and how broad the circumstances they can replicate are. If for example someone could make a self-replicating robot but the robot required boron in some critical part, its replication would be severely hampered.
Moreover, even having a self-replicating robot isn’t by itself necessarily useful if you can’t control it in detail or get them to then do exactly what you want. And a self-replicating robot with no control could be quite bad.
But these are essentially minor nitpicks, and I agree with your point if one adds the appropriate minor disclaimers.
I was thinking more in terms of the original claim. Self replicating robots able to replicate quickly enough and flexible/controllable enough to make a permanent colony on the moon for us.
I mean, I assume that was the original point instead of sending very large, slow Von Neumann machines to tile the moon with copies of themselves. That would be cool but probably not worth the expense, and it’d carry such an awful risk of backfiring on us.
Per Robin Hanson, a machine shop can put out its own mass in equipment in roughly a month or two. And yet, the economy doesn’t double every month, or even every year. Why not?
There seems to be a fair chance the reasons are mostly rooted in cognitive biases, cumulative coordination mistakes, economic rent-seeking, and so on—not anything technological.
A well planned lunar or orbital mission might well be free of these issues. Space conditions are mechanically simpler in some respects, so there’s a stronger case for pre-planning everything rather than requiring a market economy to make it work. Supporting structures are less needed, transit is less two dimensional, and solar energy can be harvested at scale with low costs in equipment density. There is also instant access to ultra-high vacuum conditions which are useful for refining. And in addition to the endless cheap sunlight, there’s no anti-nuclear lobby which can claim it’s in their back yard.
Maybe we should solve this whole not being post-scarcity problem...
If we switch the example to an excavator which outputs its own mass in an hour or two, does the answer to your question become clearer?
A quick process like that is pretty much insignificant compared to a month or two, let alone 15 years. Unless there are tens of thousands of other steps in the chain of comparable length, it doesn’t come close to explaining it.
As I see it, there are roughly four steps:
Excavating.
Refining.
Power collecting.
Manufacturing.
The ones towards the end seem to be the biggest time sinks. However, power collection should not raise it by more than a factor of two or so. I don’t think it takes many months to mine enough coal to pay for the energy costs of coal mining equipment, for example.
I think you underestimate how much work did go into e.g. Microsoft Windows, or Linux (complete with drivers, graphical environment etc etc); I meant that by operating system (not PhD project sized of course. There’s giant difference; you can call both ‘operating systems’ just as you can have mechanical engineering project of making ‘a liquid fuel rocket engine’ and it’ll be so much smaller than Space Shuttle main engine). Also I included the design of computer hardware and fab plants on the high range.
The engineering of the self replicating factory, with the exception of engineering of few components, is mostly down to building system out of components, which is more like software engineering and less like design of a space shuttle engine (or for simpler analogy a ball bearing).
Ok—what’s your evidence? My evidence:
(a) Lots of operating systems, no self-replicating synthetic hardware in commercial use.
(b) Academics build lots of operating systems, academics build at best trivial self-replicators (see e.g. http://www.news.cornell.edu/stories/may05/selfrep.ws.html)
(c) Existing natural self-replicators are very sophisticated (effectively molecular nanomachines!)
It’s untouched ground, there’s a minimum threshold for being a self replicator, but no minimum threshold for getting called ‘operating system’ (I myself can probably write an ‘operating system’ in a week or two, or even in a weekend, depending to how little we consider to be an operating system), and such replicator—I am speaking of supervised self replicator, i.e. largely under remote control—is economically ineffective compared to regular, staffed factory. A replicator is no magic—it is a moderately big complex of factories, with automatization of even the rare tasks that would normally be much much cheaper to perform with humans—there is literally no commercial incentive to design something like that, considering that it would cost billions to design. More, actually, because the hardware of the first original is going to be awfully expensive.
I must make clear what replicator I am speaking of. A practical one, which is to significant extent remotely controlled, and for which much of the difficult to manufacture electronic components (CPUs, memory chips) are shipped from Earth (they are lightweight). Such replicator would require multiple extremely heavy launches to get there, it would be very fragile, and it would grow considerably slower than even remotely wise investments. So you’ll probably be better off simply doing something else and going to moon later.
It is not really glamorous awesomeness by any means. It would largely rely on existing CNC machines and existing robotics, it would be huge, it would be messy, and it is going to break down once in a while at first (with a bit of luck that may be resolvable with remotely controlled robot, in a manner similar to trying to solder chips onto a board by holding them with chopsticks).
edit: tl;dr; it can be done, it wouldn’t be the most complex task mankind ever done, and it is going to suck.
I am calling you out—I don’t think you can write an operating system from scratch in less than a few years. (I thought tinkering in an existing toy OS for a semester in my undergrad OS class was fairly hard. I once tried writing a “real” file system from scratch—very difficult also).
How familiar are you with software engineering or mechanical engineering? You are systematically underestimating the difficulty of engineering tasks, I think. You sound like a futurist, not an engineer.
I am not sure I understand the concrete proposal you are making, nor am I sure how much “replication” should be a part of such a proposal. A lot of people are thinking very hard about economical moon/mars colonization, any thoughts on their proposals?