It makes sense, although I'm not sure what behaviours you have in mind.
Won't behaviours that tend to increase the chances of the same or similar behaviours in future tend to eventually stabilise to a fixed set of consistent behaviours, so reducing flexibility and adaptability rather than increasing it?
Some do.
Those of an asteroid do. That is why although asteroids last a long time, we don't think of them as being alive.
Other behaviors -- namely, the meta-behavior of
changing one behavior due to
some other behavior in a repeatable way, which is the physical instance of the abstraction we call a
computation -- don't. Because those behaviors are basically just switches between other behaviors. Asteroids don't do that. Hardly anything does that. Except life and the systems that make up life (and the stuff life has made ).
To formalize this:
Suppose there are two systems, S1 and S2. Suppose those systems each exhibit two behaviors at once -- call them behaviors A and B. Now suppose that for both systems, behavior A can be either X or Y. Also suppose that for system S1, B = B1, and for S2 B = B2. In other words behavior B is different between system S1 and S2.
Now suppose that in a given set of environmental states, both systems last
longer in a form that can exhibit behavior B when A = X or A = Y (depending on the environment state). That is, for environmental states { 1, 2, 3 ... N } the systems last longer in a form that exhibits B if they also exhibit A = X instead of A = Y, and vice versa for states { N +1, .... M } where M is the total number of possible environmental states.
Finally, suppose that in S1, behavior A is independent of behavior B1 but in S2 A is
dependent on B2. What can happen?
Well, clearly S1 has a 50% chance ( assuming N = 0.5 M ) to be in a state that maximizes the chances it will exhibit B1 at some point in the future. B1 might get lucky, and S1 is in the right state, then it might get lucky again, etc. That is like an asteroid sitting in space doing nothing -- if a comet is going to collide with it, and the asteroid is either moving out of the way or not moving out of the way, any other behavior in the asteroid is at the whim of the movement status. If the comet hits it, all the other behaviors that require the asteroid to remain intact will cease to be exhibited thereafter. The asteroid -- still the same "system" -- is now in the form of a million smaller mineral pieces floating in a cloud in space.
What about S2?
Well, if the dependency is such that B2 causes S2 to exhibit A = X or A = Y at the
wrong time, then S2 will have a 0% chance to be in a state that maximizes the chances it will exhibit B2 at some point in the future. In other words, behavior B2 is trying to commit behavioral suicide.
But if dependency is the opposite, B2 increases its chances of being exhibited in the future. If B2 causes S2 to exhibit A = X when A = X
maximizes the chances of B2 being exhibited in the future, it has bootstrapped itself.
And that, my friend, is how life works == billions of behaviors like B2 that lead to the continued existence of the system that can exhibit them again in the future, by
computing what the state of the system that will
maximize the chances of future existence should be.
Behaviors like B2 are examples of
computations -- the system changes behavior in some
additional way because of them. Not all computations lead to a higher chance of computation in the future -- in fact computations occur all over the place, all the time, and don't lead to anything. But when computations
do increase their chances of being repeated, special things happen.
Like life. I don't care what anyone says -- life is special, very special, because it is by far the system in the universe that exhibits the
most such behaviors. Heck, that's what life
is.
