Imagine the difficulty in designing a machine which makes silicon chips and then constructs a computer with them, given the raw materials.

Even if such a machine could be made, imagine the difficulty in redesigning it to find its own raw materials. An enormous increase in complexity.

Now imagine that you must redesign this machine yet again to require that it create copies of itself. This will require it to not only find the raw materials for making computer chips, and then making them, but also to find all the raw materials for making all its own components and then making those.

That level of complexity is almost to much to even hold in the imagination.

If it is too much for you to imagine, then stick with the vacuum tube approach or even the baseline electromagnetic relay approach.

Alas, replace "silicon chips" in my previous post with "vacuum tubes" or "electromagnetic relay" and the difficulty in imagining such an accomplishment remains.

Do you think there's something magical about human beings that makes it so that we are able to make computers but machines can't? I mean...it's not like we handcraft computer circuits, it's mostly made by machines as it is. How much of it requires human hands?

Remember, I'm not talking about design or creativity. These machines are only building something that was already designed.

It takes many humans in concert with many machines to find, extract, and manipulate the raw materials to make a computer. Proposing that one machine replace all those humans and machines and be self replicating to boot is inconceivable.

Also, humans are alive and blessed with the most versatile computer we know of, the human brain. So, yes, compared with machines, humans are seemingly magical.

It might be more practical to use an entire fleet of (specialized) machines, not only one.

At what step in the process is a human doing something that a robot couldn't be programmed to do?

I can point to the "magical" step in a thermonuclear bomb or stellar fusion which a fusion power reactor can't duplicate. The "magic" in a thermonuclear bomb is fission (of course, fission power reactors exist). The "magic" in stellar fusion is waiting billions of years for your fuel to burn, even if you did match the heat and pressure for free.
Why does it have to be just "one" machine? The basic requirement is that the machines are self-replicating. If this is best done with many different "species" of machine working in concert, then so be it.

By your standard, humans aren't self-replicating. We simply couldn't survive without the help of other species of life. We're part of a larger biological system, without which we can't survive.
The human brain might be the "magic" that can't be duplicated with programmed robots (robots that we know how to design/build). Okay, so where in the step of building a computer is the human brain necessary? What step is something that a robot can't do because it lacks the capabilities of the human brain?

Remember, I'm not talking about designing a new model of computer, I'm just talking about manufacturing an already designed computer.

Perhaps in theory a sufficiently complex machine could do what a human or any living thing does: survive long enough to grow and reproduce. I just think you are greatly underestimating the degree of complexity required and how close we are to achieving it (closer than achieving controlled nuclear fusion, you say).

I'm trying to imagine the simplest self reproducing machine. There are no requirements other than it reproduce using raw materials naturally found in its environment and that its reproductions reproduce and so on. In how much detail can you describe such a machine? Can you keep the requirements from rapidly spiraling out of control?


Concerning our side discussion:
According to this article, "Although deuterium is created at a very slow rate, it is destroyed at an extremely high rate...". It looks to me, according to the accompanying graph, that the reaction of deuterium to helium-3 takes place at more than 1000 trillion times the rate of the reaction creating the deuterium. So it seems that when you say that a fusion reactor would take billions of years to burn its fuel, you are assuming that the fuel is hydrogen, as it is in stars, when in fact the fuel will be deuterium (and perhaps tritium). The star needs to make its own deuterium, an extremely slow process, the fusion reactor does not.

Ha!

Let's create one first, with only the requirement of reproducing itself.

Suppose a self replicator could be made using only five elements. Perhaps you envision five separate machines, each specialized for finding and extracting one of those elements. They would then come together and, sharing amongst all, would have the raw materials needed for all to reproduce. If you could do that for five elements, you could do it for any number. It's coming up with the brain power that would be the biggest challenge, I think.

We have already achieved controlled nuclear fusion, as I've already noted. That's different from making a fusion power reactor.
The simplest machine may be something like this robot which is little more than a robotic arm on wheels. It uses its arm to press buttons on other machines to manufacture more of it--more or less the same way we humans build robots.

The "simplest" solution from the perspective of solving the problem may involve copying the way we humans do it. That's what we did in the auto industry. We didn't try to invent ways to manufacture cars from scratch. We took the jobs that humans were doing on the assembly line and made robots to imitate those jobs.

We humans build robots using a bunch of helper tools and machines. Thus, if we want the simplest solution for how to make a robot build other robots, a reasonable place to start is to simply imitate we humans already do it.
Note that the reaction being refered to is a deuterium+proton reaction rather than a deuterium+deuterium or DT reaction.
I'm assuming fusion is working in a star the way it works in a star, not how you might wish it to work. My point being that it's a baseless to think that there's some "continuum" of possibilities going on here when the mechanisms and physics involved are just plain different.

One way is to scale up a Tokamak. There is no longer any serious doubt that a large Tokamak could well exceed breakeven, so there's your power reactor. Sure, there are plenty of practical issues that will need to be resolved before fusion power reactors go onto the grid, but then again, there are plenty of issues that will need to be resolved for useful self replicating machines.

__________________
I say there is an invisible elf in my backyard. How do you prove that I am wrong?

Disclaimer: Avatar is not an official NASA image and does not imply any specific interplanetary or interstellar capability.

The Leif Ericson Cruiser

A machine constructing a copy of itself from materials provided by humans is not an autonomous replicator and definitely not artificial life.

I am not arguing with the prospect of building complicated networks of machines to do things. We do that already as you say. I'm arguing that autonomous, artificial life is far beyond our capabilities.

It's the self replication from raw natural resources that is the problem, not the piecing together of prefabricated parts.

Every time you say "just make a new machine to do that", you increase the complexity. I think that complexity increases exponentially until it rapidly surpasses human abilities to design such a system. We will not have the ability to create a system sufficiently complex to be called artificial life for the foreseeable future.


As for fusion:
The point is, the limiting reaction in a star is the creation of deuterium. This limiting reaction does not exist in a fusion reactor, so you can not say that since the process takes a billion years in a star, it will take a billion years in a reactor.

As I said before, there is a continuum in the variables of heat/pressure and fuel mix. You cannot say that physics does not allow certain pressures to exist, as least in the realm we are talking about for a fusion reactor. And as far as fuel mix goes, we have hydrogen, deuterium, and even tritium at our disposal. There is a continuum in the relative ratios of those elements and the right mix for a fusion reactor exists in that continuum.

There are huge technical challenges. but the process for a working fusion reactor can be calculated. There is no physical prohibition against it. I'd like to see you restate your reason for thinking it is physically impossible without using the argument that fusion in a star is a billion year long process.

It looks to me like you are proposing something like this:
A machine fabricates copies of itself from parts that are made by other machines. These other machines must also be reproduced. It takes yet more machines to provide the material for this and they also must be reproduced. It looks like a never ending, ever widening web of machines, not a one of them alive.

Machines that perform complex jobs and fabricate things, even copies of themselves if given the appropriate materials, are not alive. To call something artificial life, it must reproduce using raw materials from its environment.

Living things, artificial or not, must manipulate matter on the atomic level. A living machine must make a copy of itself atom by atom, molecule by molecule.

Humans are not alive because they can build a car. Humans are alive because they take in material from the environment and incorporated it, atom by atom, molecule by molecule, into their bodies and at some point create a replica (proto) human which in turn takes material from the environment to construct its body.

Well, that's your definition of life.

I think, within the next few decades, the definition of "life" is going to get to get very complex indeed. I would be surprised if it took more than a few decades to at least build macroscopic replicating systems that could use "raw" materials. For one concept, see Advanced Automation for Space Missions concept. Whether you want to call that life is up to you.

__________________
I say there is an invisible elf in my backyard. How do you prove that I am wrong?

Disclaimer: Avatar is not an official NASA image and does not imply any specific interplanetary or interstellar capability.

The Leif Ericson Cruiser

Automation does not equal life.

"Regeneration" as described in The Seven Pillars of Life , is a critical feature of living things. This necessarily takes place at the atomic/molecular level.

Complexity does not equal life.

Even sentience is not sufficient for a machine to be alive. A computer so complex that it achieves sentience could be made, but if it can not regenerate and reproduce, it is not alive. On the other hand, a bacterial cell which (let's assume) is not sentient, does regenerate and reproduce and is alive.

Why? A system could regenerate by building parts using bulk manufacturing processes and use those to replace broken parts. It doesn't have to manipulate each atom.

__________________
I say there is an invisible elf in my backyard. How do you prove that I am wrong?

Disclaimer: Avatar is not an official NASA image and does not imply any specific interplanetary or interstellar capability.

The Leif Ericson Cruiser

Exploring with an armada of autonomous robots

Post edit: there is no mention of self-replication in the article, but it supports the idea that fleets of specialized machines may be more useful

Is this true? What changed in the last few years to eliminate this doubt?

Is it really that easy--just scale up a Tokamak, and it'll work? How much scaled up are we talking about, here? We're talking about something bigger than ITER, right?

I'm just deducing based on the fact that if it were so easy then we'd already be doing it. Therefore, it's not "so easy", so there must be something hard about it. If we assume that there's no doubt it would work, then the hard part must be something practical--like sheer cost because it's so darn big. ITER is already so huge and expensive that it's hard to build based on funding difficulty alone. So that's why I'm deducing that you must be talking about something even bigger and more expensive.

Is that right? If not, where have I gone wrong?

Thanks--I'm honestly interested. Mainly I'm interested in the answer to the question, "How big are we talking?"

Possibly. I can envision a system that contains a computer with the information to construct itself as well as all the machinery needed to carry out that construction. But would it be life? Would it evolve? Could the potential for evolution be built into its program?

Such a system would need to be incredibly complex. Either extremely massive -- comprising the machinery of natural resource exploration, mining, and manufacturing -- or extremely miniaturized, down to the molecular level. And the question remains, would it be life, or just a very complex system of machines?

These questions are interesting and a good exercise for trying to understand what defines life. I don't know the answers. I do however think that building a machine that qualifies as artificial life is far beyond our capabilities. We are only now getting to the point where we can build life, like Frankenstein, from the parts of other life.

Beyond our current capabilities, perhaps. But in a couple of hundred years or so such artificial self-replicators might be everyday objects. This is also the sort of timescale that is required for a serious interstellar program to get underway; such an energy-intensive undertaking requires a lot of infrastructure in outer space, which we haven't got yet.

However a form of artificial, self-replicating life might be made in the next few decades; an artificial bacterium, made from entirely manufactured DNA and proteins. It might be based entirely on living prototypes, but that will change over time. Entirely artificial, living organisms will probably be the first self-replicating technology we achieve, and this achievement will give pointers towards ways of making other systems which can replicate autonomously.

__________________
New Orion's Arm Site . The Starlark . Against a Diamond Sky (OA Novella Collection) . OA Flickr set

The sort of system I would favor would be specifically programmed to NOT evolve on their own. The last thing I'd want is for the machines to evolve into Skynet.

I would consider it "life" by my definition, since I do not consider the ability to "evolve" to be a defining characteristic of "life".

The trouble is, a system which does not evolve is limited in its response to its environment. If it encounters an environment or conditions not anticipated in its design, it will fail to cope with them.

__________________
New Orion's Arm Site . The Starlark . Against a Diamond Sky (OA Novella Collection) . OA Flickr set

are carbon hydrogen oxygen & nitrogen (similiar energy bonds)

and those 7 pillars of life

i think a planet with water/ice is pretty essential colony spot

i think there are two route too artificial life (for the purpose of space colonization)

microbes to insect size (shoot a tennis ball onto europe and hope u get a good landing spot )

and something the size of a house on the larger scale of things
filled with androids and other robots
gently touch down in the location of choice and get busy or die !
(cant see anything smaller would work for mainly mechanical type life)
----------------------------------------------------
for comparison one can imagine a space ship filled with highly trained humans
all the best kit and in communication with earth or a larger mothership for extra advice, willing to give their lives to the cause of building a colony

it would be one heck of a struggle living on the margins for years as everything has to be replaced as it wears out and a sustaining system derived from that only that moons resources must be established...u may start with 20 humans (robots) and end up with 4...downshifting after the initial turbo building mode... perhaps after initial terraforming the next 100 generations are insects before getting big again as the ant colony grows ever more successful

tough but doable ....Asimo version 100 may just be capable of such a task (if we started building in that direction now and would take 50years intensive research to turn him into a founding father..and thats for mission to colonize earth or europa (easy local targets)

Whether this is a feature or a flaw depends on the desires of the creators/operators.

Hmm...I may be biased since I'm a professional computer programmer. We programmers spend most of our time stamping out unanticipated undesired behavior outside the strict bounds of exactly what we intended. So naturally, I'd want a robot to do exactly what it's supposed to do and nothing else.

Anyway, here's an interesting story:

Device Like 'Star Trek' Replicator Might Fly on Space Station


This story concerns an electron beam freeform fabrication, or EBF3 device. I find it very interesting and relevant to this discussion for many reasons.

First off, and very significantly to me, EBF3 is not just a prototyping machine but something which is apparently supposed to save production costs for aerospace components. In other words, not only is it economically viable for prototyping, but it's also economically viable for mainline production of components! On a business level, this is huge. It means much greater economic incentive and impact to develop and market EBF3 technology, compared to older prototyping technologies.

Second, EBF3 works with metal, rather than plastic or some other materials. As I've discussed in this thread, you need to build with metal if you want to build electronics. EBF3 can also mix and match different feedstocks. With one feedwire of conductive metal, and another feedwire of an insulator, you can make complex electromechanical components.

Third, EBF3 is suitable for vacuum and microgee environments. With its use of an electron beam, vacuum is a big advantage! And unlike other prototyping machines, this one has actually been tested in zero gee. That we're seriously considering actually sending one of these into space for actual use says a lot about its practicality in space applications.

This technology may prove to be a huge enabler of space colonization and space industry. It may not be necessary to have huge factories on the Moon or Mars or Earth orbit in order to manufacture components from raw resources. Relatively small EBF3 devices could do the job instead.

I seem to remember Arthur Clarke predicting this sort of vapour beam technology many decades ago; a very exciting possibility. I wonder how the strength of an electron-beam deposited metal component compares with a forged and heat-treated component (although I suppose the component could be heat-treated after deposition).

Fabrication technology on the Moon and elsewhere in space will probably be restricted to creating a relatively limited set of components- but with careful planning, that limited set of components could be combined together in a large variety of ways to produce many useful products. I doubt that many of those products will be worth exporting to Earth- they will be more useful for building up interstellar infrastructure.

__________________
New Orion's Arm Site . The Starlark . Against a Diamond Sky (OA Novella Collection) . OA Flickr set

Such machines may be autonomous and even self replicating, but I don't know how they could be called living. A hallmark of living things is that, left to their devices they do not do what you want them to, they do what is in their self interest.

It's not like a vapour beam. It's essentially a robotic welding machine, but you build entire components out of little welds instead of just using welds to fill in joints.

An electron beam is used to generate the heat for various reasons. Compared to a laser, it's more efficient, it can heat up reflective metals well, and it doesn't require eye protection. Compared to arc welding, there is no electrode erosion to worry about--good for keeping durability high and running costs low.

There are plenty of humans who do things that are not in their self interest--like overeating, or smoking, or drinking and driving...

At least when animals like ants or bees work themselves to death, it would seem to be toward some larger purpose. With many of the irrationally deadly human behaviors, there's no such excuse.

Strictly looking at this from an evolutionary point of view, another hallmark of life is variation. Characteristics which may not be in a particular individual's best interest are nevertheless present in a population because they are, in the long run, in the population's best interest. If circumstances change, characteristics that previously were detrimental to an individual may now be superior from a survival standpoint.

edit: So I should say, life behaves in the best interest of life, not the best interest of the individual. In an eat or be eaten world, every occurrence certainly can not be in the best interest of every individual.

Sure. We can look at domesticated animals as examples. Look at domesticated cows. They're really dumb and are easy fodder for predators. They just get used and slaughtered by humans. It seems like a sad life--but they live on anyway. In terms of population, domesticated cows have thrived compared to their proudly wild untamed relatives.

Same thing with domesticated corn. The features which we bred into domesticated corn mean that the darn things practically can't even sustain themselves without our help planting the seeds (wild grasses drop their seeds off the cob to distribute them naturally, but this would make them hard for humans to harvest).

The tradeoff, in the cosmic scale of things, is that being dumb domesticated beasts made for better survivability and better population levels.

The same sort of cosmic tradeoff could exist for robots. Robots which do just what they're supposed to get beyond the prototype stage and progress into full production. Robots which are buggy and "evolve" unexpected behaviors don't.