Established Member
Watching Bad Universe on DVD a couple of days ago, on one episode Phil Plait visited a facility that produced self replicating machines. A 3D printer was set up to fabricate copies of the components it was built from which could be assembled to create an exact replica of the original machine. What would happen if the process was left to run indefinitely, would you evenutally get evolution acting to constantly refine the design and if carried out long enough an "artificial" life based ecology?
Established Member
I'd say there must be some mechanism that allows for small random variations and an environment that selects from those variations. It would be interesting to see the results of such an experiment. I know artificial life is researched computationally (virtually) with e.g. cellular automata, but I don't know how complex those virtual life-forms can get. I would say to really get interesting results we need a very powerful parallel processing computer.Originally Posted by starcanuck64
Order of Kilopi
The question is: Do we want machines that can evolve? When we start building self replicating machines that start with raw materials and build themselves, so that you could, for instance, put them on an asteroid and let them build from there, I would expect we would put in a lot of error checking so that they don't evolve, along with "vitamins" (chips, or whatever) that we could cut off. After all, machines might evolve out of the resource wasteful side tasks we built them to do in the first place, and does it make any sense to allow that?
By the way, you might want to read the story "Epilogue" by Poul Anderson. Very much on subject.
I say there is an invisible elf in my backyard. How do you prove that I am wrong?
The Leif Ericson Cruiser
Established Member
Yes we do need a kind of artificial life for recycling. These machines will act similar to the complicated biological life on earth and would be quite useful to manage wastes.... I think ants, cats, mice, etc., are to some extent part of a recycling process on earth beatifully designed to look very bio natural.
Administrator
I don't think ants, cats, or mice are designed and they look natural because they are.
Established Member
In general, the recycling process is beautifully designed similar to the current recycling process designed by humans though it is naturally occuring. Perhaps we did not design it initially but as a natural process it converted to a kind of interactive process to make it fit into our life. Most probably our waste machines in future may appear similar to ants, cats, mice,...A kind of force has made humans do this process that is going to run indefinitely and automatically and no one can stop it. A kind of evolving process. I hope i have understood your point properly. I'm thinking of the natural occurances of such processes for every species and i don't know wheather animals have similar interactions with nature or not.
Order of Kilopi
Everyone is entitled to his own opinion, but not his own facts.
Order of Kilopi
No because the system doesn't have a way of gathering the raw material and, as of yet, does not self assemble.
If you did get a machine that was independent, ie it could collect the material it needed and could self assemble then maybe. More likely, in my opinion, is that it would either continue making itself perfectly and would run out of resources and wouldn't change or if there are flaws in assembly that these lines would quickly "die" out.
Evolution needs something to work on. This means things that cause changes.
I'm not saying "artificial" life couldn't be created that could evolve but the artificial life would have to be of a type that could handle changes and still function.
Established Member
Right, I left the self-sufficient aspect out.
It seems to me if you set up a durable independent self-replicating artificial system, if given enough time some of the same evolutionary forces that modified succeeding generations of biological life would work to modify artificial life. As Van Rijn points out it would be influenced by how durable the error correction was. Even multiple redundancy systems would probably allow some "mutation" in code over long periods of time. If you wanted to intentionally start an evolutionary process with independent self-replicating machines you could make the error correction less durable.
I'll check "Epilogue" out.
Last edited by starcanuck64; 2011-Oct-06 at 07:24 PM.
Order of Kilopi
Engineered systems are generally very minimalist and functional, each part highly optimized for a specific function, and the system as a whole being quite brittle in terms of tolerating variations. An error in reproduction is almost certain to produce a machine that can't function at all, or change it in a way that isn't passed on to the offspring. A machine "species" capable of evolution would have to be extremely redundant and insensitive to variations. Practically speaking, a purposely engineered system would need to be designed from the start to be capable of evolving, with a flexible genome specifying its physical construction in a way that makes random variations and combinations of genomes more likely to result in working machines.
Established Member
I was thinking that in designing an independent self-replicating machine that has the capability of surviving and reproducing in a wide range of environments you'd have to design in a fair degree of flexibility, the question being would it be enough to allow viable "offspring" from errors in reproduction.
This would be taking place on a geological time-scale allowing for many different combinations to be tried.
Order of Kilopi
Most production errors would either produce non-functional or impaired offspring, or have no effect on subsequent generations. Errors would have to occur in the software driving the machinery to produce the offspring, and software tends to be particularly brittle with respect to errors, as well as being quite easy to make extremely good at detecting and correcting errors...a corrupted program will most likely crash or produce completely non-working output, and it's not difficult to design error detection/correction into things so that the mean time between undetected errors is greater than the lifetime of the universe.
Now, you could design the machine to have a flexible genomic representation of its own structure that could be varied and combined with genetic information from other machines, or set up a bunch of self replicating machines to replicate themselves up to a point and then start running a genetic algorithm to produce self replicating offspring completely from scratch...the initial replicators basically taking the place of the tidal pools, mud pits, hydrothermal vents, and all the other places cooking up complex chemicals where life may have begun. You'd have to specifically design the machines to do this, though...just designing them to be capable of reproduction isn't enough.
Established Member
This makes sense.
I wonder what affect sentience might have on an evolutionary process involving artificial life, perhaps the machine could consciously adapt it's code to take advantage of any changes in it's genome. Then again the changes might affect it's cognitive abilities.
Posting Permissions
Reply With Quote