Randomness in Evolution: Valid and Invalid Usage

[Walter Wayne]

I see. So you regard it as a coincidence that mammals with a hundred trillion cells exist now, whereas 3 billion years ago when life started there were only single-celled organisms?

If there is no bias up or down in complexity but variation in both directions, what happens when starting from as simple as possible?

So you regard it as a coincidence that, if I have a population of bacteria some of which are resistant to penicillin and some of which aren't, when I introduce penicillin into the environment the resistant group will dominate the population after a few generations?

So you are going to characterize evolution by just one of its two characteristic subprocesses? Natural selection without mutation.

Don't be ridiculous.

These are obvious patterns, and they are very predictable. They are not random at all.

The phrase "Not that there aren't patterns," exists right in the post you replied to, followed by an example. Ridiculous?

Walt

 

[sol invictus]

You and the bulk of those who argue that evolution is non-random make similar fundamental mistakes about probability theory (viz., articulett's argument that algebra is random by the definition I gave).

Why don't you give an example of a mistake I made, mijo?

 

[Taffer]

Actually, it is you who clearly doesn't understand the very tests you performed. Yes, your data was derived from a random (i.e., equiprobable) sample, but that actual data points were most probably not uniformly distributed. To test a hypothesis, you then probably assumed (at least if you were performing a parametric test such as the z-test, t-test, or chi-squared-test) that the data took on a certain distribution and then compared that distribution to determine the probability that the means of the distributions were different purely due to the variations with within the sample.

As mentioned before, such a schema requires a much wider understanding of what "random" means than just "equiprobable", as each data point is considered to be a random variable that is described by a non-uniform probability distribution.

Uh, no mijo. The bulk of my tests were such things as maximum parsimony tree building, which uses equiprobably random distributions based on a dataset to construct a large number of hypothetical trees, then selects the most likely tree based on another criterion. Tests are then performed using a random distribution of genotypes compared to the actual dataset to determine the likelyhood of a random sample constructing the chosen tree.

 

[Taffer]

So you are going to characterize evolution by just one of its two characteristic subprocesses? Natural selection without mutation.

But they are completely seperate processes. Evolution does not require random variation to occur, only that variation exists in a given population.

 

[sol invictus]

If there is no bias up or down in complexity but variation in both directions, what happens when starting from as simple as possible?

You said "Increase in complexity over time just isn't the case." which is obviously false given the history of life on earth. Furthermore, even in some counterfactual and abstract thought experiment where you start with complex life already in existence, I see no reason to believe that increasingly complex life wouldn't evolve from there. In fact I suspect I could prove it in a mathematical model for evolution. So your original statement was wrong, and I think the position you are trying to shift to now is as well (although it might actually be interesting to discuss).

So you are going to characterize evolution by just one of its two characteristic subprocesses? Natural selection without mutation.

No, that is not the implication. Expose a population of bacteria to such a pressure and resistance to it will evolve through mutations even if it is not already present in the genome, at least if the pressure doesn't wipe out the entire population. That's predictable with essentially 100% confidence - just the opposite of random.

 

[Walter Wayne]

But they are completely seperate processes. Evolution does not require random variation to occur, only that variation exists in a given population.

We'd all still be bacteria without the random variation.

Walt

 

[Walter Wayne]

You said "Increase in complexity over time just isn't the case." which is obviously false given the history of life on earth. Furthermore, even in some counterfactual and abstract thought experiment where you start with complex life already in existence, I see no reason to believe that increasingly complex life wouldn't evolve from there. In fact I suspect I could prove it in a mathematical model for evolution. So your original statement was wrong, and I think the position you are trying to shift to now is as well (although it might actually be interesting to discuss).

In that case, there is a pattern of increased rotary phone use in North America. There were none around here prior to 1870, and now I know of several. Of course all of us are aware that rotary phones have been much more prevalent in between. In evolution was focus so much on the way back and the now that we see a trend were there isn't one.

We see a pattern because the initial condition was at one extreme, but when looking at more recent history, that trend may or may not be there. (And it is an interesting point, and it surprised me when I first read about it).

No, that is not the implication. Expose a population of bacteria to such a pressure and resistance to it will evolve through mutations even if it is not already present in the genome, at least if the pressure doesn't wipe out the entire population. That's predictable with essentially 100% confidence - just the opposite of random.

And what determines the bolding if-statement, if not whether the mutation happens in the next few generations before it is wiped out. The qualified if-statement seems to contradict your "essentially 100% confidence" comment.

Walt

 

[sol invictus]

We see a pattern because the initial condition was at one extreme, but when looking at more recent history, that trend may or may not be there. (And it is an interesting point, and it surprised me when I first read about it).

It is an interesting point. However I suspect it's wrong (that is, I suspect there is a real increase in complexity over long time periods, more or less regardless of the initial condition).

One of the really difficult questions in evolutionary biology is how to define information, or complexity as we're calling it here. I don't know how to do it - Shannon entropy of the genome is an idea, but not good enough - but I suspect there's a way, and I further suspect that if you choose a good definition you will find that evolution leads to it increasing.

The reason I say that is that I think intelligence will eventually evolve given long enough, and once that happens things change fundamentally. Natural selection of a sort will continue, but the creatures in question will be aware of the process and can manipulate it in ways that make "fitness" an increasingly irrelevant concept. As that proceeds, these creatures will improve their intelligence, build computers, write books, and generally vastly increase the information/complexity.

 

[cyborg]

We'd all still be bacteria without the random variation.

And your reasoning for this is...?

 

[mijopaalmc]

And your reasoning for this is...?

You should know since you claim to understand evolution.

 

[zosima]

I just thought I'd note.

Wolfram talks a lot about randomness in A New Kind of Science, and it is available online for free. I don't think I believe his broader conclusions, but his result that very simple systems can generate enough complexity that it appears random is indisputable and says quite a bit about the nature of the random.

I think it could do a lot to illuminate this discussion.

 

[Belz...]

You clearly don't understand probability if you think that.

Really ? Then by all means, explain to me what probability becomes if we know all the factors and variables involved (that is, no heisenberg principle applied.)

Yes, you are parroting others.

No, I'm saying the same thing they are. That's not parroting, that's agreeing. I do remember telling you it "made everything random" before I read anybody else's response to your nonsense.

You simply repeated the old canard that my definition "makes everything random". This show that you, like articulett, cyborg, and sol invictus, clearly don't understand probability theory.

So which is it ? Parroting or misunderstanding ?

I'm not claiming that evolution is necessarily random by the definition I provided; I'm claiming that people haven't presented any compelling evidence that it is deterministic random by the definition I provided.

"Deterministic random" ?

 

[Belz...]

We'd all still be bacteria without the random variation.

And your reasoning for this is...?

You should know since you claim to understand evolution.

Enlighten us. Don't play the "if you don't know, I won't tell you" routine.

 

[Dancing David]

But they are completely seperate processes. Evolution does not require random variation to occur, only that variation exists in a given population.

Thank you, so many people miss that.

 

[Dancing David]

We'd all still be bacteria without the random variation.

Walt

Whoosh.

You missed the point, natural selection can occur without random mutation.

 

[Dancing David]

It is an interesting point. However I suspect it's wrong (that is, I suspect there is a real increase in complexity over long time periods, more or less regardless of the initial condition).

One of the really difficult questions in evolutionary biology is how to define information, or complexity as we're calling it here. I don't know how to do it - Shannon entropy of the genome is an idea, but not good enough - but I suspect there's a way, and I further suspect that if you choose a good definition you will find that evolution leads to it increasing.

The reason I say that is that I think intelligence will eventually evolve given long enough, and once that happens things change fundamentally. Natural selection of a sort will continue, but the creatures in question will be aware of the process and can manipulate it in ways that make "fitness" an increasingly irrelevant concept. As that proceeds, these creatures will improve their intelligence, build computers, write books, and generally vastly increase the information/complexity.

The issue of increasing complexity is a very difficult one as you point out quite wel.. There is the 'complexity' of the structure of the organism which is what most people think about. Conversely as time pass there seems to be a limiting of genetic diversity, which makes sense.

I think that plants ability to make food from sunlight is however the coolest, and then the rest sort of follows from that. Or we would all be in the oceans. So the defintion of trends in complexity is a tough one to define. certainly there is a difference between the pre Cambrian and the post Cambrian.

 

[Dancing David]

You should know since you claim to understand evolution.

That is not debate, cease fire.

Take your nanny nanny boo boo outside.

 

[Dancing David]

I just thought I'd note.

Wolfram talks a lot about randomness in A New Kind of Science, and it is available online for free. I don't think I believe his broader conclusions, but his result that very simple systems can generate enough complexity that it appears random is indisputable and says quite a bit about the nature of the random.

I think it could do a lot to illuminate this discussion.

Cool, it is also part of chaos theory that simple rules and sensitive interactions produce determinitic but pseudo random patterns.

This point has been ignored before.

 

[cyborg]

This point has been ignored before.

It is likely to be ignored in the future.

http://www.cs.auckland.ac.nz/CDMTCS/chaitin/sciamer.html

 

[sol invictus]

Well, we did discuss chaos in the other thread. I used it to make the point that even if one were to ignore the fundamental quantum randomness we think is present in all physical interactions, macroscopic systems would still not be deterministic in any meaningful, operational sense, because of chaos.

Let me put it this way - randomness can arise essentially in two ways:

1) a "truly" random process, one whose result could never be predicted no matter how much information is available (like quantum measurements)

2) a process in which the experimenter is ignorant of some of the determining factors, making the result unpredictable (like flipping a coin, or the weather on this day next year)

My argument was that the difference between these is essentially nil in real systems, because in a chaotic system one needs to know the initial conditions with exponential precision in order to predict the behavior in the future, and that is actually impossible even in principle.

I think it might be interesting to have a thread on randomness in general. Evolution, in my opinion, is not a very interesting case, because it's quite clear how randomness enters in it (and I think nearly everyone in the thread agrees on that).