Simple definition usually is that if two organisms can't produce viable, fertile offspring then they're of different species.
Even in sexual organisms, this definition of species has some problems. You can easily have organism A that can breed with B, and B can breed with C, but A and C are incompatable. How do your divide up the species lines in this case?
In general, when a new species forms, each organism has to have others it can mate with, or else they would just die out without any offspring. The speciation process is a gradual one, and so, theoretically, there is probably a path you could follow between any two sexual organisms where any pair on the path could theoretically mate.
In Avida, for simplicity, we determine species by testing each orgasnism against the species of its parent. If it can cross-over at most points with the prototype of that species, it is marked as being part of that same species. If it cannot, we create a new species for it where it is the prototype. Not an ideal method, but it works in most cases (and we rarely need to resort to the species concept).
What's fun, is that this even works for asexual organisms. We can force all possible crossovers (in isolation of course -- this never feeds back into the system) to see if they would have any ability to mate if they has been sexual.
Dr. Charles Ofria Director, MSU Digital Evolution Lab
All that really says is that is possible for intelligent beings to start up the evolutionary process, not that this is the only way it could happen.
In Avida, we are often examining issues of how it is possible for evolution to produce organisms of greater and greater complexity. We're interested in the generic process of evolution, not necessarily bound to a specific substrate. While the system we are examining was initially designed by a human, the complexity itself is generated by the basic rules of heritable variation and selection, without a human futher interfering.
Dr. Charles Ofria Director, MSU Digital Evolution Lab
This is a very good point. Computer viruses actually have the computer as a "host" and hence fit the definition well. We tend to compare the digital organisms to computer viruses as a way of explaining them to people, but you are right that they're not the same thing.
We are, however, doing some research on viruses within Avida. Specifically, we allow organisms to inject small snippets of code into each other. Sometimes these code segments could have the ability to take over the replication mechanisms inside of the digital organisms host and force them to use up their resources to make more copies of the snippet. These are much closer to the classical definition of a virus.
Dr. Charles Ofria Director, MSU Digital Evolution Lab
Actually, you're right! Core War was the original inspiration for this entire line of research; whenever I give talks about it, that's typically where I begin ("Like all good research in Computer Science, the field of digital evolution started as a computer game...")
From that, Stein Rasmussen added mutations and developed "Core World". The programming language was still a bit too fragile, but he did see a limited amount of evolution. The first really successful program of this type was then the Tierra platform by Tom Ray.
Avida was heavily inspired by Tierra, with the main difference that we tried to set it up as a proper research platform where the user could easily control many of the parameters in the system, and record whatever data they would need. Currently there is a good amount of experimental evolution work being done around the world using Avida.
Dr. Charles Ofria Director, MSU Digitial Evolution Lab
Fortunately, I think that it'll be somewhat difficult to create a true computer virus based on this code. The Avida organisms are written in a virtual assembly language that is quite different from real-world assembly languages. The commands are simplified and designed to do *something* reasonable in just about any situation.
We've done some experiments with more complex genetic languages, but in all cases they just didn't evolve as well without very specialized mutation types.
I can think of a number of ways that it would be possible to design an evolving computer virus, but I hope they're all non-intuitive enough that we have some time before anyone manages to get one working well. I've often though about trying to extend this work into the security arena -- if I didn't have so many projects going at once right now, I'd seriously consider that.
Dr. Charles Ofria Director, MSU Digital Evolution Lab
Its generally clear that sexual reproduction has long term benefits that will help a species... genetic exchange allows multiple benefical mutations to recombine into a single organism rather than competing with each other.
But this benefit is only in the *long term*. What would allow sex to be around long enough in the first place to allow this to come into play? Any individual subgroup is likely to be more successful if they don't have to (1) find mates, (2) maintain all of the extra mechanisms to facilitate recombination, or (3) have only half of their population (the males) actually producing offspring.
There are many alternative hypotheses about how sex could get started (and in what situations it would have short-term benefits) and we're trying to explore these one-by-one in Avida.
Charles Ofria Director, MSU Digital Evolution Lab
Re:Where to get Avida
on
Digital Darwin
·
· Score: 2, Informative
To add one more comment to this (from the Michigan lab mentioned above) we'll have a Windows version of the software out "real soon now". Just this past week we've gotten everything compiled under Visual C++, and are hammering out the last few major bugs. We'll put up a beta as soon as its reasonably solid.
Evolution and Avida (from one of the authors)
on
Digital Darwin
·
· Score: 1
The parts of working on this that have amazed me most are when the evolution doesn't go as I've planned. In particular, when writing Avida, the best debugging technique I have is to just run it, and then see how some of the evolved organisms work. If I made any mistakes, they will find a way to exploit my errors.
One key thing about Avida is that its not exactly a genetic algorithm. The digital organisms must self-replicate. No matter how skilled they are at performing any of the rewarded computations, if they can't also copy themselves their genetic material will not make it into the next generation. Some people may consider this a minor difference, but it causes certain side effects, such as an evolution toward being more robust to mutations (See this previous space.com article), and in general helps prevent the population from running into a complexity barrier.
Now, back to the organisms exploiting any mistakes I've made. The story that convices most biologists that these organisms can, in some sense be considered alive is this. I was working on a project where I didn't want any more beneficial mutations to be able to occur, so that I might be able to study more ecological effects. Since I'm using a computer system, I actually have the ability to fully analyze every mutation as it happens -- I can take the resulting organism, start up a new population, and run it for a while to see how it does. Now obviously this will slow down an experiment tremendously, but if I'm willing to take such a time hit, it will work. I can then take any mutation that would be beneficial, kill off those organisms (or even just revert that mutation). I implemented it and set it up to do.
What happened? Well, I watched the run as it was going (looking particularly at these test environments), and was surprized to see the fitness of the organisms appeared to be dropping. I couldn't understand it; just because there would be no mutations to improve their ability to survive, it still shouldn't drop. So I looked more carefull in the population itself, and there the fitness appeared to continue to rise. In enither case did I see the stasis that I expected.
Upon further investigation, it turned out that the organisms had evolved a way to distinguish between the test environments and the real one. I had made a slight difference on how I gave them the input numbers to the computations they needed to be able to perform, thinking it would really matter in the end. But from the organisms persepective they were able to use this -- if the inputs looked like those from a test environment, the organisms would purposfully not perform any tasks, while if they were from a real population they would do all of the rewarded tasks they could and continue to adapt to perform more.
It shocked me that they were able to figure this out so easily. A biologist friend of mine equated it to predator avoidance -- if they showed a particular behavior they would be killed for it, so they were careful how they did it. Kind of like if a squirral wasn't careful collecting nuts, a bird might swoop down and get it. Even being careful there is an occasional problem, but they can do quite well for themselves.
I went in and fixed it: I made both the real population and the test environment as random as possible, and started my runs up again.
Did this work? Of course not. What they started to do now was just play a probability game. They would ususally do all of their tasks, but sometimes they would do none of them. If an otherwise beneficial mutation happened to occur when they weren't doing tasks, it would slip through and get into the population, never to be checked again. This really slowed down the rate of evolution (most beneficial mutations were purged), but enough still slipped though.
I am actually at a loss on how to get rid of them all! Here I have this system that should make experimental evolution all the easier because I have "complete control" over it, when in truth life does always seem to find a way.
Charles Ofria
Re:The article fails at validating evolution...
on
Digital Darwin
·
· Score: 2, Insightful
Actually, this is incorrect. We started with simple digital organisms that were only capable of self-replication, and put them in an environment where they would get more CPU time (basically have their priority increased) if they performed certain mathematical computations, in this case bitwise boolean logic operations.
The organisms were only given a nand instruction to work with, and combinations of nands (linked together carefully with proper juggling of data) could be used to create any of the others. The equals operation (returns 1 where bits in the two sequences are the same, 0 where they differ) seems to require at least 19 instructions to perform. Its important to note that we did not reward partials solutions -- they either got the correct answer of they didn't.
Indeed, this might be a bad attitude, but this is a situation where people have to continually invest money and equipment in order to provide us with a free service.
It would be a different story entirely if they were placing a limits on freedom of software. If someone isn't getting paid to do what they're doing, you shouldn't be too upset with them for making a change.
Slashdot Mirror
Simple definition usually is that if two organisms can't produce viable, fertile offspring then they're of different species.
Even in sexual organisms, this definition of species has some problems. You can easily have organism A that can breed with B, and B can breed with C, but A and C are incompatable. How do your divide up the species lines in this case?
In general, when a new species forms, each organism has to have others it can mate with, or else they would just die out without any offspring. The speciation process is a gradual one, and so, theoretically, there is probably a path you could follow between any two sexual organisms where any pair on the path could theoretically mate.
In Avida, for simplicity, we determine species by testing each orgasnism against the species of its parent. If it can cross-over at most points with the prototype of that species, it is marked as being part of that same species. If it cannot, we create a new species for it where it is the prototype. Not an ideal method, but it works in most cases (and we rarely need to resort to the species concept).
What's fun, is that this even works for asexual organisms. We can force all possible crossovers (in isolation of course -- this never feeds back into the system) to see if they would have any ability to mate if they has been sexual.
Dr. Charles Ofria
Director, MSU Digital Evolution Lab
Really? True Geekdom.... Cool!
Dr. Charles Ofria,
Director, MSU Digital Evolution Lab
All that really says is that is possible for intelligent beings to start up the evolutionary process, not that this is the only way it could happen.
In Avida, we are often examining issues of how it is possible for evolution to produce organisms of greater and greater complexity. We're interested in the generic process of evolution, not necessarily bound to a specific substrate. While the system we are examining was initially designed by a human, the complexity itself is generated by the basic rules of heritable variation and selection, without a human futher interfering.
Dr. Charles Ofria
Director, MSU Digital Evolution Lab
This is a very good point. Computer viruses actually have the computer as a "host" and hence fit the definition well. We tend to compare the digital organisms to computer viruses as a way of explaining them to people, but you are right that they're not the same thing.
We are, however, doing some research on viruses within Avida. Specifically, we allow organisms to inject small snippets of code into each other. Sometimes these code segments could have the ability to take over the replication mechanisms inside of the digital organisms host and force them to use up their resources to make more copies of the snippet. These are much closer to the classical definition of a virus.
Dr. Charles Ofria
Director, MSU Digital Evolution Lab
Woo hoo! What better sign of success could we have?
Dr. Charles Ofria
Director, MSU Digital Evolution Lab
Actually, you're right! Core War was the original inspiration for this entire line of research; whenever I give talks about it, that's typically where I begin ("Like all good research in Computer Science, the field of digital evolution started as a computer game...")
From that, Stein Rasmussen added mutations and developed "Core World". The programming language was still a bit too fragile, but he did see a limited amount of evolution. The first really successful program of this type was then the Tierra platform by Tom Ray.
Avida was heavily inspired by Tierra, with the main difference that we tried to set it up as a proper research platform where the user could easily control many of the parameters in the system, and record whatever data they would need. Currently there is a good amount of experimental evolution work being done around the world using Avida.
Dr. Charles Ofria
Director, MSU Digitial Evolution Lab
Fortunately, I think that it'll be somewhat difficult to create a true computer virus based on this code. The Avida organisms are written in a virtual assembly language that is quite different from real-world assembly languages. The commands are simplified and designed to do *something* reasonable in just about any situation.
We've done some experiments with more complex genetic languages, but in all cases they just didn't evolve as well without very specialized mutation types.
I can think of a number of ways that it would be possible to design an evolving computer virus, but I hope they're all non-intuitive enough that we have some time before anyone manages to get one working well. I've often though about trying to extend this work into the security arena -- if I didn't have so many projects going at once right now, I'd seriously consider that.
Dr. Charles Ofria
Director, MSU Digital Evolution Lab
Its generally clear that sexual reproduction has long term benefits that will help a species... genetic exchange allows multiple benefical mutations to recombine into a single organism rather than competing with each other.
But this benefit is only in the *long term*. What would allow sex to be around long enough in the first place to allow this to come into play? Any individual subgroup is likely to be more successful if they don't have to (1) find mates, (2) maintain all of the extra mechanisms to facilitate recombination, or (3) have only half of their population (the males) actually producing offspring.
There are many alternative hypotheses about how sex could get started (and in what situations it would have short-term benefits) and we're trying to explore these one-by-one in Avida.
Charles Ofria
Director, MSU Digital Evolution Lab
To add one more comment to this (from the Michigan lab mentioned above) we'll have a Windows version of the software out "real soon now". Just this past week we've gotten everything compiled under Visual C++, and are hammering out the last few major bugs. We'll put up a beta as soon as its reasonably solid.
The parts of working on this that have amazed me most are when the evolution doesn't go as I've planned. In particular, when writing Avida, the best debugging technique I have is to just run it, and then see how some of the evolved organisms work. If I made any mistakes, they will find a way to exploit my errors.
One key thing about Avida is that its not exactly a genetic algorithm. The digital organisms must self-replicate. No matter how skilled they are at performing any of the rewarded computations, if they can't also copy themselves their genetic material will not make it into the next generation. Some people may consider this a minor difference, but it causes certain side effects, such as an evolution toward being more robust to mutations (See this previous space.com article), and in general helps prevent the population from running into a complexity barrier.
Now, back to the organisms exploiting any mistakes I've made. The story that convices most biologists that these organisms can, in some sense be considered alive is this. I was working on a project where I didn't want any more beneficial mutations to be able to occur, so that I might be able to study more ecological effects. Since I'm using a computer system, I actually have the ability to fully analyze every mutation as it happens -- I can take the resulting organism, start up a new population, and run it for a while to see how it does. Now obviously this will slow down an experiment tremendously, but if I'm willing to take such a time hit, it will work. I can then take any mutation that would be beneficial, kill off those organisms (or even just revert that mutation). I implemented it and set it up to do.
What happened? Well, I watched the run as it was going (looking particularly at these test environments), and was surprized to see the fitness of the organisms appeared to be dropping. I couldn't understand it; just because there would be no mutations to improve their ability to survive, it still shouldn't drop. So I looked more carefull in the population itself, and there the fitness appeared to continue to rise. In enither case did I see the stasis that I expected.
Upon further investigation, it turned out that the organisms had evolved a way to distinguish between the test environments and the real one. I had made a slight difference on how I gave them the input numbers to the computations they needed to be able to perform, thinking it would really matter in the end. But from the organisms persepective they were able to use this -- if the inputs looked like those from a test environment, the organisms would purposfully not perform any tasks, while if they were from a real population they would do all of the rewarded tasks they could and continue to adapt to perform more.
It shocked me that they were able to figure this out so easily. A biologist friend of mine equated it to predator avoidance -- if they showed a particular behavior they would be killed for it, so they were careful how they did it. Kind of like if a squirral wasn't careful collecting nuts, a bird might swoop down and get it. Even being careful there is an occasional problem, but they can do quite well for themselves.
I went in and fixed it: I made both the real population and the test environment as random as possible, and started my runs up again.
Did this work? Of course not. What they started to do now was just play a probability game. They would ususally do all of their tasks, but sometimes they would do none of them. If an otherwise beneficial mutation happened to occur when they weren't doing tasks, it would slip through and get into the population, never to be checked again. This really slowed down the rate of evolution (most beneficial mutations were purged), but enough still slipped though.
I am actually at a loss on how to get rid of them all! Here I have this system that should make experimental evolution all the easier because I have "complete control" over it, when in truth life does always seem to find a way.
Charles Ofria
Actually, this is incorrect. We started with simple digital organisms that were only capable of self-replication, and put them in an environment where they would get more CPU time (basically have their priority increased) if they performed certain mathematical computations, in this case bitwise boolean logic operations.
The organisms were only given a nand instruction to work with, and combinations of nands (linked together carefully with proper juggling of data) could be used to create any of the others. The equals operation (returns 1 where bits in the two sequences are the same, 0 where they differ) seems to require at least 19 instructions to perform. Its important to note that we did not reward partials solutions -- they either got the correct answer of they didn't.
For those interested, we also have a lot of our data related to the paper on our servers, http://myxo.css.msu.edu/papers/nature2003
Charles Ofria (Second author of the paper, and primary author of the software)
Indeed, this might be a bad attitude, but this is a situation where people have to continually invest money and equipment in order to provide us with a free service.
It would be a different story entirely if they were placing a limits on freedom of software. If someone isn't getting paid to do what they're doing, you shouldn't be too upset with them for making a change.