(no subject)
Lenski, Ofria, Pennock, and Adami, "The Evolutionary Origin of Complex Features."
One of the classical questions regarding evolutionary theory is how complex features (such as the eye) develop, given the difficulty of imagining "intermediary" forms which would possess useful qualities; multi-step developments would be impossible, it could be argued, when intermediary steps would be selected against. This paper addresses these questions using a computer simulation of evolution, called Avida, in which "digital organisms," self-replicating programs running on a virtual machine, evolve "complex features" in which they derive "energy" required for reproduction by performing logical operations. Such operations require as many as a few dozen instructions to execute, implying that direct mutations from a simple ancestor to a operation-performing organism would be extremely unlikely; nonetheless, organisms are observed to evolve in mutliple steps to be able to perform logical operations, and organisms which perform all available rewarded operations consistently came to dominate the population. Very complex operations are evolved only in circumstances where less complex operations are rewarded, though specific precursor operations are not required.
However, given all of that, the paper doesn't reveal anything particularly unexpected; just as one would expect, multi-step evolution of these complex features is found to involve several intermediary steps, many of which are detrimental but which replicate "long enough" regardless. The paper is silent as to exactly how much of a "leap of faith" of this sort evolution can be expected to take, or how the impact of intermediate steps (detrimental mutations leading to advantageous ones, or simpler functions leading to more complex ones) in digital organisms relates to real ones; is the much greater complexity of the features desired for real organisms offset by greater population sizes, more generations, larger genomes, or etc?
Also, its interesting to note that the "ceiling" of the evolution modelled here was the most complex rewarded logical function. I wonder if a general schematic could be made for an infinite progression of increasingly complex logical functions, associated with increasing levels of rewards; if applied to an Avida simulation, could this result in unbounded evolution (in the domain of the schematic)? I would expect someone to have tried this.
Ofria, Wilke, "Avida: A Software Platform for Research in Computational Evolutionary Biology."
This paper is simply a description of the Avida program and how it can be used. In terms of general concepts there's not much to say about this paper, beyond the basic idea of Avida which I outlined above. The system does seem to be very flexible (allowing a range of rulesets and environmental conditions), so if I were to do an AL project on evolutionary effects (such as the idea I mentioned above) it would be a reasonable way of doing it. My biggest concerns with the program is the lack of inter-organism interaction (though they mention this as an expected feature, and may have implemented it since this article was written) and also lack of organism death conditions (organisms would currently only die when overwritten or after executing a given number of operations, depending on experiment settings.) These omissions greatly impact the possible dynamics in an experiment (particularly, at least in the second case, the dynamics of the of of the organisms' evolution.)
One of the classical questions regarding evolutionary theory is how complex features (such as the eye) develop, given the difficulty of imagining "intermediary" forms which would possess useful qualities; multi-step developments would be impossible, it could be argued, when intermediary steps would be selected against. This paper addresses these questions using a computer simulation of evolution, called Avida, in which "digital organisms," self-replicating programs running on a virtual machine, evolve "complex features" in which they derive "energy" required for reproduction by performing logical operations. Such operations require as many as a few dozen instructions to execute, implying that direct mutations from a simple ancestor to a operation-performing organism would be extremely unlikely; nonetheless, organisms are observed to evolve in mutliple steps to be able to perform logical operations, and organisms which perform all available rewarded operations consistently came to dominate the population. Very complex operations are evolved only in circumstances where less complex operations are rewarded, though specific precursor operations are not required.
However, given all of that, the paper doesn't reveal anything particularly unexpected; just as one would expect, multi-step evolution of these complex features is found to involve several intermediary steps, many of which are detrimental but which replicate "long enough" regardless. The paper is silent as to exactly how much of a "leap of faith" of this sort evolution can be expected to take, or how the impact of intermediate steps (detrimental mutations leading to advantageous ones, or simpler functions leading to more complex ones) in digital organisms relates to real ones; is the much greater complexity of the features desired for real organisms offset by greater population sizes, more generations, larger genomes, or etc?
Also, its interesting to note that the "ceiling" of the evolution modelled here was the most complex rewarded logical function. I wonder if a general schematic could be made for an infinite progression of increasingly complex logical functions, associated with increasing levels of rewards; if applied to an Avida simulation, could this result in unbounded evolution (in the domain of the schematic)? I would expect someone to have tried this.
Ofria, Wilke, "Avida: A Software Platform for Research in Computational Evolutionary Biology."
This paper is simply a description of the Avida program and how it can be used. In terms of general concepts there's not much to say about this paper, beyond the basic idea of Avida which I outlined above. The system does seem to be very flexible (allowing a range of rulesets and environmental conditions), so if I were to do an AL project on evolutionary effects (such as the idea I mentioned above) it would be a reasonable way of doing it. My biggest concerns with the program is the lack of inter-organism interaction (though they mention this as an expected feature, and may have implemented it since this article was written) and also lack of organism death conditions (organisms would currently only die when overwritten or after executing a given number of operations, depending on experiment settings.) These omissions greatly impact the possible dynamics in an experiment (particularly, at least in the second case, the dynamics of the of of the organisms' evolution.)