INTRODUCTION
In a paper (1) published in Nature in 2003, researchers (acronym LOPA after their surnames), conducted an experiment with digital organisms interacting in an environment conducive to Darwinian style evolution. The organisms:-
1) reproduce,
2) compete for resources,
3) mutate.
Point 3) is important in that it causes variation and in conjunction with 1) ensures differential fitness when it comes to competing for resources 2).
The experiment was performed in an attempt to gain insight concerning the origin of complex features. The report covered just one aspect of this evolution, namely the ability of one system of organisms to evolve the complex logical function, EQU, (equals, as in the mathematical =), out a set of 9 other possible functions, none of which were in the original population.
This essay deals very briefly with that paper, describing the digital organisms, the experiment, and the results. A much larger one I posted at another site can be found at:-
http://www.theologyweb.com/campus/t109039
Unfortunately an awful lot of detail has had to be left out, including many of the results. So I suggest you examine the larger essay. Much better still, the original paper - given that all the detail is there.
A DIGITAL ORGANISM?
In this particular case, the digital organism (DO) was constructed in Avida, a freely available system developed at Michigan State University, and used for research into DOs. A brief description of Avida can be found at reference (2).
The Avida DOs are somewhat like independent little software computers, running around within a computers memory. Each has a circular genome which can grow in length, and is made up of computer instructions of which there are 26 different types, including instructions which allow the genome to copy itself (i.e. reproduce).
During reproduction, random mutations can occur. These can be point mutations where one instruction is replaced by another (each with equal probability), deletions, or insertions. Other kinds of mutations can occur, e.g. duplications.
Instructions are things like nand, IO, push, pop, nop, sub, The first two are important. Nand is an instruction that does have an associated logical function NAND. The function NAND can be built up using as few as 5 instructions, including a nand. Other functions such as AND, OR, EQU used more instructions including more nands. EQU for example, required 5 nands included in a total of 19 instructions as a minimum, so the researchers thought. IO (input/output) was important because if it followed any nand instruction, then a result could be checked to see if one of the logical functions had been performed that is, if it had evolved. If so an appropriate energy reward was supplied.
References (3), (4) and (5) provide some reference to digital logic gates/functions.
THE EXPERIMENT
LOPA, fed Avida with 3,600 identical DOs that could replicate but perform no logic functions. Each had a genome of 50 instructions. Each obtained the same amount of energy which was used to execute its genomic instructions, including reproduction.
During reproduction, random errors could occur such as:-
1) Point mutations (error rate 0.0025 errors per instruction copied).
2) Insertions and deletions of instructions (error rate 0.05 per genome copied).
A point mutation replaced one instruction with another. Each instruction had an equal chance of being replaced by any other instruction.
Most mutations were harmful or neutral. Only a few were beneficial in that they increased fitness.
A mutation (point, insertion, or deletion) changed the genome and possibly altered the phenotype, that is, the DOs ability to replicate, extract and use energy, or its robustness. Thus, these mutations led to differences between organisms and hence their reproductive success.
Selection depended on the effect of the mutation on the genome itself (e.g. the organism could die, or be very clunky), and on the relationship of the organism to its environment. The researchers did not influence selection. The organism and its environment worked it out.
The organisms competed for energy needed to execute instructions. By executing its instructions a DO could (if it was of the right type) obtain more energy to carry on instruction execution as well as copy itself.
Energy could be obtained in one of two ways:-
1) It was given in units the longer the genome, the more units it received.
2) It was given at an increased rate of units if the organism performed a logic function (AND, EQU, NAND, AND, ), something the founding organisms could not do.
These logic functions were performed on 32 bit strings contained in two registers, with the result going into a third. They could be built up by evolving instruction sequences which included nand instructions, such that combinations of the nands in the context of the other instructions, caused one of these logic functions to be performed. For example, the sequence of instructions nand, nop, IO, , caused the BX can CX registers to be naned, with the result going to the AX register. The IO caused the AX register to be output which could then be checked. If it was found that the NAND function had been performed, then the organisms rate of energy uptake would be increased by a factor of 2.
The reason behind this is as follows. NAND is a simple logic function. The shortest version of it could be built from 5 instructions only, which, obviously, included one nand instruction. Because of this simplicity, the reward for a DO evolving it would be to have its energy uptake increased by a factor of 2. AND is a bit more complex giving a reward factor of 4. XOR is better still its factor was 16. EQU topped the list at 32. EQU needed at least 5 nand instructions in association with a lot of other code, which the authors estimated to be 19 instructions in all, at a bare minimum.
An organism could perform any or all of the nine possible logic functions during its lifetime but no extra energy was obtained by repeatedly performing the same function. No single mutation in the original population could produce a logic function. Rather, several mutations had to appear in the same lineage, and they had to be coordinately executed in order to perform even the simplest of functions.
THE RESULTS. THEY WERE INTRIGUING.
LOPA concentrated their report on the evolution of the most complex function (EQU), in one population only. However, many different populations actually evolved this EQU function.
In the particular study case they found that, although the initial population started with some 3600 organisms, the solution to the genotype exhibiting the stable EQU function was on a path some 344 generations long. This short distance was so because as time went on, increasing numbers of organisms had evolved upon which selection could operate. This dominant genome had increased its instructions from 50 to 83 and it could perform all 9 functions to obtain energy. EQU actually appeared after 111 generations even though most mutations within the overall population were harmful or neutral. The authors were not surprised because the path to this dominant genome was the path linking the winners in each generation.
But it was the role of the deleterious mutations that were the most surprising.
Of the harmful mutations that had appeared in this winning genome path, most were only slightly deleterious. It appeared that they were able to hitchhike from generation to generation, on the backs of beneficial mutations. But there were two mutations that were nearly catastrophic. They caused a >50% drop in fitness. However, they did not wipe out the organism. One (it affected reproduction) was nullified by a subsequent mutation at a distant site that interacted with it. The other, which occurred in two individuals of the winning line knocked out a simple NAND function. Yet it was shown that this mutation was needed for the EQU to evolve in the very next step..
LOPA were able to show the mutual dependency of many mutations on each other, as well as the dependence of all functions on certain instructions and mutated instructions. Thus, the first DO to evolve EQU had 60 instructions and 35 of these were required. Eliminate any and the organism would fail. They determined that some mutated instructions were so important that they were retained for all logical functions, including the simpler ones. A few instructions, five, were retained from the ancestral population. Three of these were important in replication.
They demonstrated that the evolution of EQU depended on its ability to build up from simpler and different functions, something Darwin had predicted in his Origin of Species, and confirmed repeatedly by subsequent observations.
It appeared that EQU was a very fragile function, unlikely to last. Its functioning could easily be destroyed by a mutation. Yet as the authors noted, it did last. Once it evolved, it persisted because it was so valuable (look at its energy gain), that any subsequent defective mutants were eliminated by selection.
And so the report of LOPA continued analyzing, classifying, and describing. They even noted that one DO had evolved the EQU with 17 instructions. Remember, they determined that 19 would be the shortest!
The researchers tried various experiments to test the effects of different selective environments on the ability to evolve EQU. For example, one or two of the simpler logic functions would not be rewarded. All other conditions remained constant. In all environments, at least one population evolved EQU.
In the 36 environments tested, 34% of the populations evolved EQU, only slightly less than the reward all functions environment as discussed above.
THE OLD TORNADO THROUGH THE JUNK YARD AND THE 747 AGAIN
LOPA note that, given a genome length of 50 and 26 instructions, that there are ~5.6x10[sup]70[/sup] (or (26[sup]50[/sup]) different genotypes. However, this is an underestimate, given that genome length increases through evolution.
Given this, and given a particular winning genotype, what do you think the chances are of it evolving?
One chance in 56,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000
Therefore these digital organisms could not have evolved not even within their own environment? Or is there something wrong with the tornado through the junk yard analogy as it is often applied by creationists?
LOPA CONCLUDE
They note that one big advantage of studying digital organisms, is that exact genealogies can be traced and examined because there are no missing links. They were able to show the exact dependence of the evolution of a complex function on simpler functions being used as stepping stones. This was a prediction made by Darwin and is something that is seen elsewhere by studying the evolution of living organisms. They write:-
They go on to state some of the differences between DOs and living organisms then conclude by noting how such simulations can provide insight on some evolutionary problems that have, to date, been somewhat intractable.
MY CONCLUSION
If you refuse to accept the empirical evidence for evolution, to the point of laughing at it, or dismissing it as unworthy of consideration even, then perhaps this report on digital organisms may give some small pause for thought?
Clearly the odds against the winning genotype evolving were humongously low. Yet it evolved.
If you think Darwin or ToE to be silly, then how do you explain it?
Regards, Roland
REFERENCES
(1) Richard E. Lenski, Charles Ofria, Robert T. Pennock & Christoph Adaml, The evolutionary origin of complex features, Nature, 423, 8-May-2003, p 139-144.
(2) Wicki reference- http://en.wikipedia.org/wiki/Avida
(3) Logic gates/functions http://hyperphysics.phy-astr.gsu.edu/hbase/electronic/gate.html#c1
(4) Why the nand is important http://hyperphysics.phy-astr.gsu.edu/hbase/electronic/nand.html#c1
(5) From Wicki:- http://en.wikipedia.org/wiki/Digital_circuit About 4/5 of the way down is a bit more on the universality of nand/NAND. That is, the others can be constructed from it.
In a paper (1) published in Nature in 2003, researchers (acronym LOPA after their surnames), conducted an experiment with digital organisms interacting in an environment conducive to Darwinian style evolution. The organisms:-
1) reproduce,
2) compete for resources,
3) mutate.
Point 3) is important in that it causes variation and in conjunction with 1) ensures differential fitness when it comes to competing for resources 2).
The experiment was performed in an attempt to gain insight concerning the origin of complex features. The report covered just one aspect of this evolution, namely the ability of one system of organisms to evolve the complex logical function, EQU, (equals, as in the mathematical =), out a set of 9 other possible functions, none of which were in the original population.
This essay deals very briefly with that paper, describing the digital organisms, the experiment, and the results. A much larger one I posted at another site can be found at:-
http://www.theologyweb.com/campus/t109039
Unfortunately an awful lot of detail has had to be left out, including many of the results. So I suggest you examine the larger essay. Much better still, the original paper - given that all the detail is there.
A DIGITAL ORGANISM?
In this particular case, the digital organism (DO) was constructed in Avida, a freely available system developed at Michigan State University, and used for research into DOs. A brief description of Avida can be found at reference (2).
The Avida DOs are somewhat like independent little software computers, running around within a computers memory. Each has a circular genome which can grow in length, and is made up of computer instructions of which there are 26 different types, including instructions which allow the genome to copy itself (i.e. reproduce).
During reproduction, random mutations can occur. These can be point mutations where one instruction is replaced by another (each with equal probability), deletions, or insertions. Other kinds of mutations can occur, e.g. duplications.
Instructions are things like nand, IO, push, pop, nop, sub, The first two are important. Nand is an instruction that does have an associated logical function NAND. The function NAND can be built up using as few as 5 instructions, including a nand. Other functions such as AND, OR, EQU used more instructions including more nands. EQU for example, required 5 nands included in a total of 19 instructions as a minimum, so the researchers thought. IO (input/output) was important because if it followed any nand instruction, then a result could be checked to see if one of the logical functions had been performed that is, if it had evolved. If so an appropriate energy reward was supplied.
References (3), (4) and (5) provide some reference to digital logic gates/functions.
THE EXPERIMENT
LOPA, fed Avida with 3,600 identical DOs that could replicate but perform no logic functions. Each had a genome of 50 instructions. Each obtained the same amount of energy which was used to execute its genomic instructions, including reproduction.
During reproduction, random errors could occur such as:-
1) Point mutations (error rate 0.0025 errors per instruction copied).
2) Insertions and deletions of instructions (error rate 0.05 per genome copied).
A point mutation replaced one instruction with another. Each instruction had an equal chance of being replaced by any other instruction.
Most mutations were harmful or neutral. Only a few were beneficial in that they increased fitness.
A mutation (point, insertion, or deletion) changed the genome and possibly altered the phenotype, that is, the DOs ability to replicate, extract and use energy, or its robustness. Thus, these mutations led to differences between organisms and hence their reproductive success.
Selection depended on the effect of the mutation on the genome itself (e.g. the organism could die, or be very clunky), and on the relationship of the organism to its environment. The researchers did not influence selection. The organism and its environment worked it out.
The organisms competed for energy needed to execute instructions. By executing its instructions a DO could (if it was of the right type) obtain more energy to carry on instruction execution as well as copy itself.
Energy could be obtained in one of two ways:-
1) It was given in units the longer the genome, the more units it received.
2) It was given at an increased rate of units if the organism performed a logic function (AND, EQU, NAND, AND, ), something the founding organisms could not do.
These logic functions were performed on 32 bit strings contained in two registers, with the result going into a third. They could be built up by evolving instruction sequences which included nand instructions, such that combinations of the nands in the context of the other instructions, caused one of these logic functions to be performed. For example, the sequence of instructions nand, nop, IO, , caused the BX can CX registers to be naned, with the result going to the AX register. The IO caused the AX register to be output which could then be checked. If it was found that the NAND function had been performed, then the organisms rate of energy uptake would be increased by a factor of 2.
The reason behind this is as follows. NAND is a simple logic function. The shortest version of it could be built from 5 instructions only, which, obviously, included one nand instruction. Because of this simplicity, the reward for a DO evolving it would be to have its energy uptake increased by a factor of 2. AND is a bit more complex giving a reward factor of 4. XOR is better still its factor was 16. EQU topped the list at 32. EQU needed at least 5 nand instructions in association with a lot of other code, which the authors estimated to be 19 instructions in all, at a bare minimum.
An organism could perform any or all of the nine possible logic functions during its lifetime but no extra energy was obtained by repeatedly performing the same function. No single mutation in the original population could produce a logic function. Rather, several mutations had to appear in the same lineage, and they had to be coordinately executed in order to perform even the simplest of functions.
THE RESULTS. THEY WERE INTRIGUING.
LOPA concentrated their report on the evolution of the most complex function (EQU), in one population only. However, many different populations actually evolved this EQU function.
In the particular study case they found that, although the initial population started with some 3600 organisms, the solution to the genotype exhibiting the stable EQU function was on a path some 344 generations long. This short distance was so because as time went on, increasing numbers of organisms had evolved upon which selection could operate. This dominant genome had increased its instructions from 50 to 83 and it could perform all 9 functions to obtain energy. EQU actually appeared after 111 generations even though most mutations within the overall population were harmful or neutral. The authors were not surprised because the path to this dominant genome was the path linking the winners in each generation.
But it was the role of the deleterious mutations that were the most surprising.
Of the harmful mutations that had appeared in this winning genome path, most were only slightly deleterious. It appeared that they were able to hitchhike from generation to generation, on the backs of beneficial mutations. But there were two mutations that were nearly catastrophic. They caused a >50% drop in fitness. However, they did not wipe out the organism. One (it affected reproduction) was nullified by a subsequent mutation at a distant site that interacted with it. The other, which occurred in two individuals of the winning line knocked out a simple NAND function. Yet it was shown that this mutation was needed for the EQU to evolve in the very next step..
LOPA were able to show the mutual dependency of many mutations on each other, as well as the dependence of all functions on certain instructions and mutated instructions. Thus, the first DO to evolve EQU had 60 instructions and 35 of these were required. Eliminate any and the organism would fail. They determined that some mutated instructions were so important that they were retained for all logical functions, including the simpler ones. A few instructions, five, were retained from the ancestral population. Three of these were important in replication.
They demonstrated that the evolution of EQU depended on its ability to build up from simpler and different functions, something Darwin had predicted in his Origin of Species, and confirmed repeatedly by subsequent observations.
It appeared that EQU was a very fragile function, unlikely to last. Its functioning could easily be destroyed by a mutation. Yet as the authors noted, it did last. Once it evolved, it persisted because it was so valuable (look at its energy gain), that any subsequent defective mutants were eliminated by selection.
And so the report of LOPA continued analyzing, classifying, and describing. They even noted that one DO had evolved the EQU with 17 instructions. Remember, they determined that 19 would be the shortest!
The researchers tried various experiments to test the effects of different selective environments on the ability to evolve EQU. For example, one or two of the simpler logic functions would not be rewarded. All other conditions remained constant. In all environments, at least one population evolved EQU.
In the 36 environments tested, 34% of the populations evolved EQU, only slightly less than the reward all functions environment as discussed above.
THE OLD TORNADO THROUGH THE JUNK YARD AND THE 747 AGAIN
LOPA note that, given a genome length of 50 and 26 instructions, that there are ~5.6x10[sup]70[/sup] (or (26[sup]50[/sup]) different genotypes. However, this is an underestimate, given that genome length increases through evolution.
Given this, and given a particular winning genotype, what do you think the chances are of it evolving?
One chance in 56,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000
Therefore these digital organisms could not have evolved not even within their own environment? Or is there something wrong with the tornado through the junk yard analogy as it is often applied by creationists?
LOPA CONCLUDE
They note that one big advantage of studying digital organisms, is that exact genealogies can be traced and examined because there are no missing links. They were able to show the exact dependence of the evolution of a complex function on simpler functions being used as stepping stones. This was a prediction made by Darwin and is something that is seen elsewhere by studying the evolution of living organisms. They write:-
LOPA (ref 1) below. Highlighting is mine. said:
They go on to state some of the differences between DOs and living organisms then conclude by noting how such simulations can provide insight on some evolutionary problems that have, to date, been somewhat intractable.
MY CONCLUSION
If you refuse to accept the empirical evidence for evolution, to the point of laughing at it, or dismissing it as unworthy of consideration even, then perhaps this report on digital organisms may give some small pause for thought?
Clearly the odds against the winning genotype evolving were humongously low. Yet it evolved.
If you think Darwin or ToE to be silly, then how do you explain it?
Regards, Roland
REFERENCES
(1) Richard E. Lenski, Charles Ofria, Robert T. Pennock & Christoph Adaml, The evolutionary origin of complex features, Nature, 423, 8-May-2003, p 139-144.
(2) Wicki reference- http://en.wikipedia.org/wiki/Avida
(3) Logic gates/functions http://hyperphysics.phy-astr.gsu.edu/hbase/electronic/gate.html#c1
(4) Why the nand is important http://hyperphysics.phy-astr.gsu.edu/hbase/electronic/nand.html#c1
(5) From Wicki:- http://en.wikipedia.org/wiki/Digital_circuit About 4/5 of the way down is a bit more on the universality of nand/NAND. That is, the others can be constructed from it.
