Unstructured thoughts - Radio emerges from the electronic soup

Radio emerges from the electronic soup [New Scientist]

For me the main interest about this story is that it shows how "You get what you select for."

The experimenters set up the system so that an oscillating output was selected for. They did not select for the creation of an actual oscillator. Evolving an oscillator is just one way that a system like this can give an oscillating output.

Evolutionary design processes can be very sensitive to external conditions, even to the point that a successful design at one longitude might fail at another one because of the effects of the subtle duration/temperature differences. When we embrace evolutionary design, we also embrace non-linearity, so it's a powerful but often unpredictable approach.

Of course, we can include stability or predictability into our selection criteria. In the physical world this puts a lot of pressure on the imagination of the designers of the experiment, and involves having to deliberately expose the system to a set of stressors, singly and in combination. The more critical the system, the more thorough this process need to be (as with non-evolutionary approaches, too).

It's not too different for software, but software suffers from fewer "field effects" - it's more like a mechanical constructor set than a radio constructor set. Software has defined connections between its parts. It should be possible to determine if a particular module is directly or indirectly connected with (and so can be affected by) another module. In the physical world, this is much more difficult, as the experimenters were reminded by the oscillator experiment linked to above.

Perhaps the contrast is not so clear cut, as there are a number of media in a computer through which field effects can propagate between modules or bundles of modules - the most obvious one being the operating system. OS's can be (and despite appearances, are) designed to minimise these effects. Modern OS's try to make each application feel like it's the only one running, other than the components it needs to do it's job.

The software runs on a computer, and this exists in the real world (even an emulator runs on a real computer), so physical effects can affect its running. Most of the time this minimised by very robustly designed hardware components (CPU and memory chips, printed circuits etc.) that are very resistant to errors when run within their designed conditions.

Modern PC hardware is a physical system that has been designed through an (informal but truly) evolutionary approach, with reliability of code execution, cost, size, mass, power consumption, speed and all of those other criteria that you see listed in PC magazines (and some that aren't) being the attributes selected for. My guess is that a PC designer would look at the New Scientist story and look skyward, eyebrows arching, lips pursing. "Been there, done that".

One benefit of the (formally) evolutionary approach is that it has no preconceptions. Like a person who has no preconceptions, this could lead to moments of brilliance and also to moments of intense social awkwardness. The typically short generation time of virtual experimental systems allow avenues to be explored that would be shunned in the real world, and give them a chance to bear fruit. If these systems need to be translated into the real world the design must include all of the expected environment stressors.

One factor that has been selected for in bridge and building design is for the designer not to end up in prison, or sued into the ground, and this has had a clear effect of structural engineering and design. (They are also quite keen on not killing people).

Funny to think of a lawyer as an environmental stressor, isn't it?