On Learning to Trust the Unexpected: Toward a Systematic Apologetic for Evolvable Hardware

Evolvable hardware (EH) has been described as a composition of reconfigurable hardware and a learning algorithm that optimizes configurations according to predefined operational requirements. EH, to be most effective, requires its users to surrender significant amounts of design authority to an automated process. Even though EH has shown significant promise in creating novel designs, working engineers often show reluctance in relying on them. Understanding the causes of such reluctance is a necessary step toward constructing cogent arguments for the adoption of EH methods in practice. In this paper, we attempt to examine some of the roots of observed reluctance and construct some preliminary arguments to counter it