The strategy being advocated in this proposal is to let natural selection do most of the work of directing evolution and producing complex software. This software will be ``wild'', living free in the digital biodiversity reserve. In order to reap the rewards, and create useful applications, we will need to domesticate some of the wild digital organisms, much as our ancestors began domesticating the ancestors of dogs and corn thousands of years ago.
The process must begin with observation. Digital naturalists must explore the digital jungle, observing and publishing on the natural history, ecology, evolution, behavior, physiology, morphology, and other aspects of the biology of the life forms of the digital ecosystem. Much of this work will be academic, like the work of modern day tropical biologists exploring our organic jungles (which I have been doing for twenty years).
However, occasionally, these digital biologists will spot an interesting information process for which they see an application. At this point, some individuals will be captured and brought into laboratories for closer study, and farms for breeding. Sometimes, breeding may be used in combination with genetic engineering (insertion of hand written code, or code transferred from other digital organisms). The objective will be to enhance the performance of the process for which there is an application, while diminishing unruly wild behavior. Some digital organisms will domesticate better than others, as is true for organic organisms (alligators don't domesticate, yet we can still ranch them for their hides).
Once a digital organism has been bred and/or genetically engineered to the point that it is ready to function as an application for end users, they will probably need to be neutered to prevent them from proliferating inappropriately. Also, they will be used in environments free from the mutations that will be imposed on the code living in the reserve. By controlling reproduction and preventing mutation, their evolution will be prevented at the site of the end user. Also the non-replicating interpreted virtual code, might be translated into code that could execute directly on host machines in order to speed their operation.
The organisms living in the biodiversity reserve will essentially be in the public domain. Anyone willing to make the effort can observe them and attempt to domesticate them. However the process of observation, domestication and genetic engineering of digital organisms will require the development of much new technology. This is where private enterprise can get involved. The captured, domesticated, engineered and neutered software that is delivered to the end user will be a salable product, with the profits going to the enterprise that made the efforts to bring the software from the digital reserve to the market.
It seems obvious that organisms evolving in the network-based biodiversity reserve will develop adaptations for effective navigation of the net. This suggests that the most obvious realm of application for these organisms would be as autonomous network agents. It would be much less likely that this kind of evolution could generate software for control of robots, or voice or image recognition, since network based organisms would not normally be exposed to the relevant information flows. Yet at this point we surely can not conceive of where evolution in the digital domain will lead, so we must remain observant, imaginative in our interpretations of their capabilities, and open to new application possibilities.