Evolutionary Algorithms and ID

You’re right. It was so obvious for me that I omitted to mention this point. Okay. Let me rephrase it: « the function of VDJ recombination is to produce a large repertoire of antibodies with different binding specificities ». Is it OK?

What does that mean?

1 Like

Oh no, not this again!

There is the well known 2nd Law of Thermodynamics, but there is no 2nd Law of Information Theory. There is an equivalent statement for Relative Information, which boils down to expelling excess randomness to be “relatively closer” to some given pattern. There is a long thread discussing it … (I’ll look).

The movements of grains of sand blown by the wind are very much random - contingent, if you like - yet these patterns are common. Why isn’t a designer required??

1 Like

I suspect that you also didn’t bother to do the calculation before claiming without any reservation that @gpuccio had already demonstrated this to be true.

Not even close. Let’s talk about one VDJ recombination event and one function: a catalytic antibody.

Since we know that we can get the target, at least in some cases, from searching only 10^8 clones from an unimmunized library, the FI = -log2(10^-8) = 26. That’s the FI for a single heavy or light chain in a single recombination event that is random with respect to fitness.

How many of these occur over how much time?

3 Likes

I didn’t say that - please watch your quoting.

This question is maybe for @Giltil, but perhaps @Mercer and others might pipe in too. I’m am woefully ignorant of these information-based arguments, I’m used to regular old thermodynamic entropy and such, so please bear with me.

My question is, is this “500 bits” criteria (which at this point seems completely arbitrary to me, but that’s probably for another post) for intelligent design a cumulative thing or for each individual step? So if @Mercer is calculating 26 bits of FI in a single recombination event, will 20 recombination events “make it” or would we have to find a single “thing” that has 500 bits of FI?

Despite the complexity of the language and information theory, it looks to this naive outsider, like it’s the old “slow accumulation of small changes” vs “leaps of change at specific times” argument. Am I wrong?

2 Likes

No

1 Like

You are correct, not wrong.

1 Like

The answer is that 20 events will NOT « make it » and we would have to find a single « thing » that has 500 bits of FI.

1 Like

Neither AVIDA nor Tierra are GA’s. All GA’s are goal oriented. The generational solutions are guided towards the goal.

When someone designs a computer program to do something and it does it, it does so by design.

GA’s have the specifications and criteria the solution has to meet. And all generational solutions are guided towards that target.

Edited! Discourse lost the attribution somewhere along the line.

The one I wrote on Tuesday wasn’t.

You can keep making this claim, but it is false, it always has been false, and it always will be false.

3 Likes

@Roy, Look up the word “algorithm”- clearly you have no idea what it means.

How did anyone determine that the immune system evolved by blind, mindless and purposeless processes? If you can demonstrate that then you don’t even have to consider what the immune system produces.

FI is function specific. If the single “thing” is a function then you are correct.

1 Like

I have that topic in mind, just haven’t had time to do it right. Soon … :slight_smile:

No, you have no idea what it means.

There are algorithms to generate random values. If you think generating random data counts as a goal, then you can go ahead and claim that GAs are all goal-oriented, but that’ll destroy the rest of your argument.

This is also false.

In most GAs there is no criteria the solution has to meet, only criteria for preferring one ‘organism’ over another. You’d know this if you’d ever written one, which is why I think you haven’t.

3 Likes

That’s that silly and roundabout way to claim everthing humans touch must be designed. It is entirely possible for programs to run without the designer of the program having much, if any idea about what will result from the running of the program.

In fact some times programmers make programs exactly because they don’t know what they will do, and are interested in making and running their programs as a substitute for real-world experiments. Because they want to understand why the things happen the way they do? Because the number of interactions are so numerous, and so sensitive to local conditions, that it is impossible to predict what they will do. They are asking basic, in fact somewhat naive questions like “What happens if we do this?”

If you want to claim that the products of a program like that were “designed”, then that’s a rather vacuous idea of design you have in mind. Design by unknowing and unpredicted accident.

It would be asinine to claim that what those programs then do has been “designed”. When people say something was designed they normally mean someone knew what they were trying to make, that what resulted was intended to result, and they were deliberately working towards that exact goal(or at least a set of such similar results).

GA’s have the specifications and criteria the solution has to meet. And all generational solutions are guided towards that target.

It is entirely possible to have GAs without targets. Here’s one: http://boxcar2d.com/
The programmer has made an evaluation in setting up the GA which is that moving further to the right is better than not. But there is no specific target set, and the exact solution this results in after some X number of iterations isn’t known beforehand. In fact the algorithm can be executed to an arbitrary number of generations, provided you have enough time and your computer has enough memory. Since it can be executed to an arbitrary number of generations, it is meaningless to say that the GA has any target, since what it results in heavily depends on how much time it is given and what the initial conditions and parameters are set to. It is not possible to predict beforehand what solution it comes up with except empirically, by re-running the algorithm many many times over and over again.

Is the algorithm designed? Yes! Are the particular solutions that evolve designed? No, not under any normal or sensible definition of design. That’d be like saying I “designed the outcome” of me tossing a handful of dice on the ground. After all, I picked up those dice and tossed them.

Sorry, you’re being silly with how you are trying to claim credit for design, by demanding that even unintended, unknown, and unpredicted byproducts of our actions should also count as “designed”.

6 Likes

algorithm- a procedure for solving a mathematical problem (as of finding the greatest common divisor) in a finite number of steps that frequently involves repetition of an operation

GA’s are search heuristics, Roy. Blind, mindless and purposeless processes do not search.