building intelligent structure with random evolution requires series of non useful steps

Question

In order to build an intelligent design such as a bridge or a mechanical watch, you need a series of non useful steps.

As far as I understand, in random evolution theory, the "intelligence" involved is the process of natural selection.

The problem with this is that natural selection requires that each step be advantageous right away. If I want to build a bridge, I need many steps which would be disadvantageous until the final product which would be advantageous. This would be like trying to write a computer program where each line of code must be advantageous to the program. An intelligent design just cannot be done like this.

How does modern biology deal with this problem? To say it all happened at once, is not possible since the probabilities against this are simply outside the capabilities of a random system. Furthermore even if by some freak monumental event it happened, it would need to happen again many times since there is a huge diversity of intelligent designs in life forms such as lungs in birds vs gills in fish, or sexless reproduction in bacteria versus male/female reproduction in animals (which are fundamentally different).

(Also, as side question has there been ANY experiment demonstrating that such a thing is even possible? I mean for example, to have a computer simulation of random self-replicating molecules, then having an intelligent human being picking out the best version and eliminating the others, then repeating, etc. until you have an intelligent design.)

Answer 1

What you say is true, but the answer as to how modern evolutionary theory understand this considers other ideas about living things which seem to allow so called 'neutral traits'.

Neutral traits being those phenotypes which seem to have no particular relationship to fitness. Toungue rolling is an often mentioned example. Some folks argue that there is only adaptive changes and phenotype, others like Gould and Lewontin argue that some traits can be the result of pleiotropy.

So first of all this sort of discussion can tell us that the question is argued both ways in a convincing enough way that it is not resolved in biology - some argue quite confidently that all traits are selected for and things like flagellar motors and upright ambulation and eyeballs evolve because of and experiencing constant evolutionary pressure. Others argue that these genes that produce these things are doing something else useful while they are becoming these later more useful traits.

The common twist here is plieotropy. That the genes that are producing the phenotype one is observing is also participating in one or more other biological traits that may is being positively selected for.

A more interesting example of how this works is the domesticated silver fox experiment. Foxes which have had domestication bred into them (just to show that it can be done by selection) lose their red color and their ears are softer and fuzzier. Nobody tried to select for these traits, but they come associated with each fox that is domesticated. Are these traits plieotropic or only very tightly linked? So far they have not been exhibited separately. In any case I think its fair to say that the experiment didn't care about the color of the fox - only its meekness towards people. These traits come along for the ride.

So it might be with any other desirable genetic trait. If a gene becomes closer to producing a lung from a gill, if that gene is sticking around it must be because there is positive selection for that trait, even if the lung like gill is not particularly useful for that fish.

Sorry but I actually disagree that it is obvious that non-useful steps must be had for any such complicated trait. If that were true then you would have a point, but not every biologist would give that to you.

It might be a bit surprising that such discussions are and continue to be matters of faith in science, but I hope you don't dismiss it because of this either. All sciences do assume some things that make their view of the world make more sense. But there are also other components of these theories which are consistently observed and understood that demand them... until they don't.

I don't give really any credence whatsoever when someone says 'this is a widely accepted theory by the scientific community'. This in itself is not a scientific statement, but more the hope that someone else has one somewhere. But a refutation must also take under consideration what is being refuted.

Answer 2

"This would be like trying to write a computer program where each line of code must be advantageous to the program. An intelligent design just cannot be done like this."

On the contrary, this is precisely how most successful programs are developed nowadays: using incremental test-driven design. You start off with a blank program and a test (typically "does it compile") which the blank program fails but which it can be made to pass by a trivial amendment (such as adding a function called "main"). With that success under your belt you move on to another test - e.g. "does it produce output", which might be satisfied by the addition of a "print 'dddagcahx'" statement.

Suppose we've decided that the purpose of our shiny new program is to take a year and tell us whether or not it is a leap year. Then our next four tests, expressed in the sort of language commonly used for such tests, might be

    Assert.That(Year.IsLeap(1996), Is.True);
    Assert.That(Year.IsLeap(1997), Is.False);
    Assert.That(Year.IsLeap(1900), Is.False);
    Assert.That(Year.IsLeap(2400), Is.True);

Now it might seem that the next step would involve a "big bang" conversion of our "print random nonsense" program into a fully fledged leap year calculator, but in fact we can pass two of the new tests simply by always outputting "True" (or alternatively always outputting "False") regardless of what input we are given. So we change "print 'dddagcahx'" to "print 'True'" (or print "False") and already we're passing four out of six tests. Even a stopped clock is right twice a day.

And so on. Even the largest program can be developed incrementally in this manner. Every small addition to the program is "advantageous" in that it enables the program to pass tests which it previously failed.

It used to be that programs were not developed in this way. Back in the 80's, say, the prevailing philosophy was "Big Design Up Front". Not a line of code would be written until after months or even years of meticulous design had taken place. The big problem was that this typically didn't work. Millions would be spent on software projects which then had to be scrapped because of design flaws which only became apparent during implementation - or, worse, only when the program was in the hands of customers. This could have been averted if there had been a partially-functioning system at every stage of the development process.

So the message of software development is that an intelligent design CAN be done like this, i.e. incrementally and "where each line of code must be advantageous to the program"; and it is the alternative "it all happens at once" approach which is problematic for reasons which you yourself have pointed out.

(See e.g. exercism.io for an excellent hands-on introduction to test-driven development.)