Michael Conrad, Klaus-Peter Zauner
Biological systems possess enviable information processing abilities, which are rooted in the self-organization of context-sensitive building blocks. Molecular computing can utilize this principle. Our objective in the present chapter is to show that this opens up a realm of information processing that is inaccessible to programmable machines. Our second objective is to present a tabletop prototype that illustrates a methodology for pursuing this direction.
Algorithmic complexity theory provides a framework for elucidating the comparative capabilities of programmable and nonprogrammable systems. Programmable architectures are amenable to a more compressible description, concomitant to the fact that they must conform to a simple user manual. To implement complex input- output behavior, it is necessary to supply a complex program. The programmer therefore must be the source of complexity. Biomolecular architectures are sharply different: Complexity is inherent. The capabilities are constructed by orchestrating a repertoire of complex components through an adaptive process. The number of functions that can be implemented is limited by the time available for adaptation and may not be larger than that in programmable systems. In this chapter, we will argue that the complexity of the actual achievable behavior is greater.
John von Neumann (1951) referred to such noncompressible complexity in a discussion of the visual cortex:
It is not at all certain that in this domain a real object might not constitute the simplest description of itself, that is, any attempt to describe it by the usual literary or formal-logical method may lead to something less manageable and more involved. (1951, 24)
In our case, the real objects are proteins. We will show that it is possible to utilize the conformational dynamics of proteins to process input signal patterns—though at this stage not in a manner that transcends formal description.