Von Neumann probes

[kStrout]

just don't have any variation of any kind in your machines, or they will evolve and take over.

[Camacha]

just don't have any variation of any kind in your machines, or they will evolve and take over.

Or, without variation, have them not evolve and still take over

[magnemoe]

That's my point: it is not true. To woodwork a spoon or cabinet on traditional machines, you need a properly filled workshop. With modern CNC equipment, you could do with one machine if you wanted. Yes, it is a more complex machine, but like I said, also much more versatile.

There is a reason digital manufacturing and printing has taken the industrial world by storm: you can produce much, much more products at a somewhat reduced (but not unusable slow) production speed. Only true mass products are still manufactured in 'dumb' production lines, everything else (including books, bridges and bikeframes) is done in more flexible digital production. Assuming the probes have some sort of self improving evolutionary development mechanism, digital manufacturing is the way to go.

Yes, also note that modern manufacturing is about cost and product quality. CNC is nice as you can make one off machine parts much cheaper than the tradition way.

Computer assisted tools are as you say very nice with limited production volumes.

However an von-neuman machine at least an interstellar probe would be magnitudes more clunky, time is pretty irrelevant it used 200 years getting to the target, if it uses 100 years making an copy nobody will mind.

Today manufacturing is all about cost and efficiency, scrap that.

Fun real world story, car mechanics are unable to work on farm or construction machinery, main issue in mentality, in an garage you have all the tools you need, any part you order comes the next day, the owner want the car back the same condition it was and expect it to work for 25.000 km afterwards.

On the other hand, you have to fix an combiner, speed is important its an storm coming in, however you might jury rig as mad as it only has to last some days, after the harvest is in you have a year to fix it.

For construction equipment things can get an magnitude more intense speed might be even more important and money is rarely an issue.

Its require an totally different way of thinking.

[Camacha]

Yes, also note that modern manufacturing is about cost and product quality. CNC is nice as you can make one off machine parts much cheaper than the tradition way.

Computer assisted tools are as you say very nice with limited production volumes.

However an von-neuman machine at least an interstellar probe would be magnitudes more clunky, time is pretty irrelevant it used 200 years getting to the target, if it uses 100 years making an copy nobody will mind.

A lot of time on your hands, but less space and machinery to produce what you need is exactly what makes digital production suitable. It is less quick than a full blown production line, but you can produce many, many times the number of different parts on the same 'line'. It is useless to fly a full production facility with little to no flexibility in space, but taking a small assortment of machines that together can produce anything (and limitless variation of that anything) is just the thing you need.

Also remember that Von Neumann machines do not really need massive production capabilities. Exponential growth will provide all the mass production capabilities you could ever desire. All you need is a bit of time, of which you have plenty.

Fun real world story, car mechanics are unable to work on farm or construction machinery, main issue in mentality, in an garage you have all the tools you need, any part you order comes the next day, the owner want the car back the same condition it was and expect it to work for 25.000 km afterwards.

On the other hand, you have to fix an combiner, speed is important its an storm coming in, however you might jury rig as mad as it only has to last some days, after the harvest is in you have a year to fix it.

For construction equipment things can get an magnitude more intense speed might be even more important and money is rarely an issue.

Its require an totally different way of thinking.

I am not sure how that story is relevant

[Jouni]

That's my point: it is not true. To woodwork a spoon or cabinet on traditional machines, you need a properly filled workshop. With modern CNC equipment, you could do with one machine if you wanted. Yes, it is a more complex machine, but like I said, also much more versatile.

An my point was that replicating a von Neumann machine is more like building the CNC equipment than making a spoon with it. That requires several branches of industry and much more equipment than building (relatively) simple machines like the Saturn V.

[Camacha]

An my point was that replicating a von Neumann machine is more like building the CNC equipment than making a spoon with it. That requires several branches of industry and much more equipment than building (relatively) simple machines like the Saturn V.

You do not necessarily need more machines for more complexity. Look at a humble plastic printer. That little machine can print thousands of different parts, for which you would have needed a vast factory full of complex injection moulding, rotary casting and all kinds of different plastic production machines. Not to mention all the support equipment, as moulds and injection machines do not make themselves. Now, you can make do pretty much the same with just one compact and cheap machine. That machine also requires a number of different technologies and devices to be produced itself, but nowhere near what you would need for a factory full of more traditional machines.

That is before even taking some form of natural selection into account - if you let machines work out what is most efficient themselves, you can be sure a lot of excess weight is cut out very quickly. Maybe some tools or resources are communally shared, who knows.

[Jouni]

You do not necessarily need more machines for more complexity. Look at a humble plastic printer. That little machine can print thousands of different parts, for which you would have needed a vast factory full of complex injection moulding, rotary casting and all kinds of different plastic production machines. Not to mention all the support equipment, as moulds and injection machines do not make themselves. Now, you can make do pretty much the same with just one compact and cheap machine. That machine also requires a number of different technologies and devices to be produced itself, but nowhere near what you would need for a factory full of more traditional machines.

That little plastic printer contains a computer at its core. The chips in the computer come from a semiconductor fab, which typically costs billions to build and contains hundreds of large special-purpose machines. The printer is made of hundreds or perhaps even thousands of materials, and the number of different chemicals and materials used for producing those materials is even higher. Then just think about the amount of equipment required for producing those chemicals and materials, and for mining the raw materials used in the production. Building that "humble plastic printer" probably depends on a nontrivial fraction of our entire civilization.

Everything just keeps getting more and more complex all the time.

[andrewas]

The thing is, we know how to build every component of that printer. We can list all the machines needed, all the resources we need. Do that for all the components in all the machines needed to produce the printer, and you eventually wind up with a set of machines that can replicate the entire collection.

That's the basis for a von Neumann machine. It doesn't need any magical technology, the hardest part will be automating the whole process and creating a control AI sophisticated enough to recover from problems, adapting to shortages of certain resources, and deal with everything that is normally left to a human crew.

[Jouni]

The thing is, we know how to build every component of that printer. We can list all the machines needed, all the resources we need. Do that for all the components in all the machines needed to produce the printer, and you eventually wind up with a set of machines that can replicate the entire collection.

Let's assume that we build a von Neumann probe like that. It's probably huge, and carries a lot of equipment that no longer exists on Earth. We launch it towards a promising solar system...

...and hope that it finds an exact duplicate of the Earth. Otherwise it's in trouble, because it may not find all the raw materials it needs, or have the right equipment for processing them. Now it needs to improvise and build completely new machines for tasks nobody had even thought about before. It's just a single probe, so it may not be able to figure out everything before it runs out of some critical resource, or before making a critical mistake.

Maybe it's better to launch many probes to the same destination. That way, the mission isn't doomed, if one individual isn't creative enough or makes a bad choice. Unfortunately, even that may not be enough. The probes may come up with a solution, but they may be unable to build the machines the solution involves quickly enough.

Perhaps the real solution is to seed a self-sustaining ecosystem at the destination, giving the probes enough time to adapt to the conditions and to figure out new ways to do things. But how do we do that, if we don't even know what the destination is like before the launch? The obvious solution is to stay in contact with the source, and launch resupply missions once the destination has been properly explored. At this point, the whole thing starts feeling more like colonization than von Neumann machines.

[Ralathon]

Let's assume that we build a von Neumann probe like that. It's probably huge, and carries a lot of equipment that no longer exists on Earth. We launch it towards a promising solar system...

...and hope that it finds an exact duplicate of the Earth. Otherwise it's in trouble, because it may not find all the raw materials it needs, or have the right equipment for processing them. Now it needs to improvise and build completely new machines for tasks nobody had even thought about before. It's just a single probe, so it may not be able to figure out everything before it runs out of some critical resource, or before making a critical mistake.

Maybe it's better to launch many probes to the same destination. That way, the mission isn't doomed, if one individual isn't creative enough or makes a bad choice. Unfortunately, even that may not be enough. The probes may come up with a solution, but they may be unable to build the machines the solution involves quickly enough.

Perhaps the real solution is to seed a self-sustaining ecosystem at the destination, giving the probes enough time to adapt to the conditions and to figure out new ways to do things. But how do we do that, if we don't even know what the destination is like before the launch? The obvious solution is to stay in contact with the source, and launch resupply missions once the destination has been properly explored. At this point, the whole thing starts feeling more like colonization than von Neumann machines.

Even if we manage to build a magic von neumann machine that perfectly replicates itself in any situation, we'd still be in heavy contact with the thing. The entire point behind a Von Neumann swarm is to get a lot of data for a relatively small cost. Guiding the first few generations until the swarm is sufficiently large would likely be standard procedure. After those first few replications the swarm is so big that a single failed replication is no longer mission critical, and we can afford to just watch the data roll in.

Also, I heavily contest your point that you always need a more complex machine to build another. Going by that logic you would never have a rise in complexity, as the child can never be more complex than the parent. Yet we see the exact opposite in nature and technology. We've build our current technological wonderland from sticks, stones and fire. Sure, it took us a few thousand years, but it is a clear sign that you can use simple machines to make more complex ones. The same thing can be observed in nature: 2 billion years ago we had pond scum, now we have cheetas and killer whales. If you give a probe a good 3d printer, an energy source, some resource extraction device and an arm to assemble stuff, it could rebuild itself. It would take a really long time, and probably take loads of intermittent steps, but provided your programming is smart enough it could be done.

[Jouni]

Also, I heavily contest your point that you always need a more complex machine to build another. Going by that logic you would never have a rise in complexity, as the child can never be more complex than the parent.

The rise in complexity is something nobody really understands. At a high level, we can talk about evolution, but we still don't know how adaptive processes transform entropy into information. Multiple independent agents, random chance, and reinforcement of both successes and failures are probably all essential.

Machines can only produce machines that are simpler than themselves, but a system consisting of many simple machines can produce a machine more complex than any individual machine. Through trial and error, mistakes and death, the entire system slowly gains complexity. As the system grows more complex, it can produce and support more complex machines. The system itself is a machine, so it can't produce another system as complex as itself. It can split into multiple independent systems of lesser complexity, and each of them can eventually grow as complex as the original, as long as the conditions remain favorable.

So, essentially:

• If you launch one von Neumann probe to a destination, it tries to improvise, makes mistakes, and dies.
  
• If you launch multiple von Neumann probes to a destination, they try to improvise, but lack the means to seed a self-sufficient system that's complex enough to produce more probes.
  
• If you launch colonization probes to a destination, and integrate the colony into the source system, the entire system grows in complexity. Eventually the colony subsystem becomes complex enough that you can separate it as a self-sufficient system and start producing more probes.
  

That gives three possible solutions to the Fermi paradox, depending on how many new colonies a colony can seed off, on the average, before running out of resources or making a fatal mistake:

1. If the number is more than 1, the galaxy is full of advanced civilizations. The only reason we haven't seen them is that they deliberately stay hidden from us.
   
2. If the number is around 1, the most advanced civilizations in the galaxy only span a handful of star systems. They survive for long periods of time, but they keep moving from system to system, as the old systems eventually become uninhabitable.
   
3. If the number is less than 1, technological civilizations are short-lived and rare. Eventually someone makes a fatal mistake, and the entire civilization dies.
   

[andrewas]

You keep asserting that a machine cannot create a machine if equal or greater complexity, but have not provided any evidence of that. To my understanding, the only limit on the complexity of the product is the data storage of the creating machine - and we can easily store a complete description of a machine with less complexity than the machine itself.

[Jouni]

You keep asserting that a machine cannot create a machine if equal or greater complexity, but have not provided any evidence of that. To my understanding, the only limit on the complexity of the product is the data storage of the creating machine - and we can easily store a complete description of a machine with less complexity than the machine itself.

The evidence is every machine that has ever been built.

On the other hand, there is no evidence beyond science fiction stories that a machine could create something more complex than itself.

[Bill Phil]

The evidence is every machine that has ever been built.

On the other hand, there is no evidence beyond science fiction stories that a machine could create something more complex than itself.

What about a machine increasing it's complexity using some of the resources in a star system? Then it could create copies of the original...

And what about cells? When they duplicate, the complexity is changed very little.

[KSK]

You keep asserting that a machine cannot create a machine if equal or greater complexity, but have not provided any evidence of that. To my understanding, the only limit on the complexity of the product is the data storage of the creating machine - and we can easily store a complete description of a machine with less complexity than the machine itself.

Agreed. It's a trite example but you could regard a cell as nothing more than an exceptionally complicated machine. Given the right raw materials, a cell has no problem creating other cells of equal complexity. Similarly, a viable von Neumann machine would probably need to be manufactureable from as few raw materials as possible, which I suspect is one reason why nobody has built one yet. However that's just a failure of current technology rather than some ironclad law that a machine cannot build a machine of equal complexity or greater.

Machines can only produce machines that are simpler than themselves, but a system consisting of many simple machines can produce a machine more complex than any individual machine.

This is just wordplay. What's the difference between a complex machine and a system of simple machines? I would argue 'nothing at all' apart from the fact that making them different supports your argument. It's equivalent to saying that I can't build a working internal combustion engine but I can build a system of simpler parts (cams, gears, pistons, cranks) that will power a car.

The rise in complexity is something nobody really understands. At a high level, we can talk about evolution, but we still don't know how adaptive processes transform entropy into information.

True - we don't know the fine details behind this rise in complexity but the basic principle isn't hard. Evolution is blind. It's a series of step by step changes with each change being independent of the preceding changes. There's a vanishingly low probability of 'walking back' any of those steps, even if that would result in a simpler outcome.

Put another way, there's a higher probability of complexity increasing than complexity decreasing.

Edited August 6, 2015 by KSK

[Jouni]

What about a machine increasing it's complexity using some of the resources in a star system? Then it could create copies of the original...

This has been discussed over the past few pages. My answer is in the topmost message on this page.

And what about cells? When they duplicate, the complexity is changed very little.

This was discussed earlier in this thread. Cells only duplicate under favorable conditions. Those conditions only occur in complex ecosystems or in artificial environments. In either case, the duplicating cells represent only a small fraction of the overall complexity of the system.

- - - Updated - - -

This is just wordplay. What's the difference between a complex machine and a system of simple machines? I would argue 'nothing at all' apart from the fact that making them different supports your argument. It's equivalent to saying that I can't build a working internal combustion engine but I can build a system of simpler parts (cams, gears, pistons, cranks) that will power a car.

As I said, a complex machine (a system of simple machines) can produce a simple machine (a machine that is bit more complex than the individual machines in the system). If you can produce the parts of an engine, you can probably assemble them into an engine. On the other hand, if you are the internal combustion engine yourself, you probably can't produce the parts without outside help.

- - - Updated - - -

True - we don't know the fine details behind this rise in complexity but the basic principle isn't hard.

The basic principle is the hardest part. We understand the low-level details with independent agents, feedback loops, and random events quite well, but we don't understand how the high-level behavior emerges from that.

[KSK]

As I said, a complex machine (a system of simple machines) can produce a simple machine (a machine that is bit more complex than the individual machines in the system). If you can produce the parts of an engine, you can probably assemble them into an engine. On the other hand, if you are the internal combustion engine yourself, you probably can't produce the parts without outside help.

Fine. So the 'complex machine' builds a set of simpler machines, each a little bit more complex than the individual parts of the 'complex machine.' The 'complex machine' then assembles those incrementally more complex parts into a new machine. That new machine is composed of more complex parts than the original 'complex machine'.

So there you have it - a complex machine that has just built a more complex machine.

The basic principle is the hardest part. We understand the low-level details with independent agents, feedback loops, and random events quite well, but we don't understand how the high-level behavior emerges from that.

No - the basic principle (or low level detail) is the easy part. As you said, we understand that. We also understand how that can lead to high level behaviour. What we can't do is predict that high level behaviour ahead of time.

Edited August 6, 2015 by KSK

[ZetaX]

This was discussed earlier in this thread. Cells only duplicate under favorable conditions. Those conditions only occur in complex ecosystems or in artificial environments. In either case, the duplicating cells represent only a small fraction of the overall complexity of the system.

If this were correct, then abiogenesis would have been impossible. A long time ago there indeed probably was some kind of (pre)cell that lived and procreated only using the chemicals that were on earth.

[Jouni]

Fine. So the 'complex machine' builds a set of simpler machines, each a little bit more complex than the individual parts of the 'complex machine.' The 'complex machine' then assembles those incrementally more complex parts into a new machine. That new machine is composed of more complex parts than the original 'complex machine'.

Now you're assuming that the "complex machine" is capable of assembling those "more complex parts" into a "new machine". Hence the "complex machine" must be more complex than in the original scenario, where it was only capable of producing "more complex parts", but not machines that are significantly more complex than those parts.

No - the basic principle (or low level detail) is the easy part. As you said, we understand that. We also understand how that can lead to high level behaviour. What we can't do is predict that high level behaviour ahead of time.

Understanding requires more than just stories that are at least superficially consistent with the evidence. If you can't make accurate predictions or produce other tangible results, you don't really understand the phenomenon at all.

- - - Updated - - -

If this were correct, then abiogenesis would have been impossible. A long time ago there indeed probably was some kind of (pre)cell that lived and procreated only using the chemicals that were on earth.

And that probably happened incrementally in a chemical system that was much more complex than the individual protocells (or whatever was around back then).

[ZetaX]

And that probably happened incrementally in a chemical system that was much more complex than the individual protocells (or whatever was around back then).

There had to be some point back when complexity was less, otherwise earth would have existed since forever. Take that point in time.

If you actually consider a planet formed "naturally" as an artificial object or an ecosystem, then my previous point still applies: you seem to consider everything an ecosystem, thus rendering your statement trivially true, but also worthless.

[Jouni]

There had to be some point back when complexity was less, otherwise earth would have existed since forever. Take that point in time.

If you actually consider a planet formed "naturally" as an artificial object or an ecosystem, then my previous point still applies: you seem to consider everything an ecosystem, thus rendering your statement trivially true, but also worthless.

This branch of the discussion just repeats what was already said earlier in this thread. Once a system becomes self-sustaining, it can grow in size and complexity over time, as long as the conditions remain favorable. This is completely different from self-replicating systems, however.

[ZetaX]

Why is a cell in a "naturally formed" ocean that recreates itself not violating your claim¿ We are assuming there not being any other cells (yet), so nothing else there is an "ecosystem". I am not even talking about what happens to it, only that this cell can fully copy itself many times.

[Jouni]

Why is a cell in a "naturally formed" ocean that recreates itself not violating your claim¿ We are assuming there not being any other cells (yet), so nothing else there is an "ecosystem". I am not even talking about what happens to it, only that this cell can fully copy itself many times.

Because the cell doesn't duplicate itself in an ocean. It duplicates itself in a complex self-sustaining system of chemical reactions. (Note that the term 'ecosystem' is often used for non-biological systems with similar characteristics as biological ecosystems.)

[ZetaX]

Because the cell doesn't duplicate itself in an ocean. It duplicates itself in a complex self-sustaining system of chemical reactions. (Note that the term 'ecosystem' is often used for non-biological systems with similar characteristics as biological ecosystems.)

No, that system of reactions must not be self-sustaining. They are simply there already. It might use them until none are left (then they probably die).

Again: your setting is pointless as you in the end are counting everything as an ecosystem, thus devoiding your argument from any meaning. If we actually build such probes, you could still say that "it is just the universe sustaining a complex ecosystem".

[KSK]

Now you're assuming that the "complex machine" is capable of assembling those "more complex parts" into a "new machine". Hence the "complex machine" must be more complex than in the original scenario, where it was only capable of producing "more complex parts", but not machines that are significantly more complex than those parts.

Yes I am but that doesn't make any difference to my argument. Start with a complex machine that is a) composed of simpler sub-machines, capable of building slightly more complex sub-machines than it, itself is made from and c) capable of assembling those more complex sub-machines into a new machine. Watch that complex machine build a more complex machine.

Understanding requires more than just stories that are at least superficially consistent with the evidence. If you can't make accurate predictions or produce other tangible results, you don't really understand the phenomenon at all.

Incorrect. Go and read up on computable vs non computable problems. It is entirely possible to have a process that you completely understand and yet the only way to predict what the process will do is to run it and see what happens.