A thermodynamically closed system that is far from equilibrium can increase the amount of physical design provided it is either front loaded or has an intelligent agent (like a human) within it.
A simple example: A human on a large nuclear powered space ship can write software and compose music or many other designs. The space ship is closed but far from equilibrium. But complexity can still increase because of the human intelligent agent.
Consider then a robot whose sole purpose is to make other robots like it or even unlike it in a similarly thermodynamically closed system. It can do this provided the software is front loaded into the robot.
Can the robot make something more irreducibly complex than itself in such a thermodynamically closed environment? I’d say, “YES”, but in a qualified way, it can evolve it provided that it is front loaded with the goal of making robots with more IC than itself.
Simple illustration, write a piece of software that can duplicate itself, it then essentially a software robot. Make the software such that each generation of software must go through a useless Rube Goldberg ritual like processing a set of randomly generated short passwords (say 3 characters). What do I mean? The first generation would look like:
String password_1 = "ABC";
if ( password_1.equals( "ABC" ) ) {
// proceed with replication....
}
With each generation, the robot lineage is pre-programmed to make each offspring increase the number of passwords it must make in order to procreate. Generation 2 would have software like:
String password_1 = "ABC";
String password_2 = "123";
if ( password_1.equals( "ABC")
&& password_2.equals ("123")
)
{
// proceed with replication....
}
Again this addition of useless complexity to each generation. By useless, I mean Rube Goldberg complexity.
Thus by the millionth generation, the robot must process a million passwords in order to procreate the next generation, whereas the first generation only had to process 1 password.
The irreducible complexity in the robot is substantially higher than the first generation. One can see this could be analogous to increase in IC in biology if the strategy of increasing Rube Goldberg complexity was a front loaded goal.
Do I believe this is how complexity evolved on the Earth? Not most of it, maybe some of it at best. I believe genetic entropy dominates, but I just put this idea on the table for consideration.
NOTES
I use the definition of closed systems from
http://www.bluffton.edu/~bergerd/nsc_111/thermo2.html