Software hardening-unifying software reliability strategies

Semantic reliability of multi-agent intelligent systems

“…Since the belief structures get updated in an IA, in effect, an IA can never be completely tested. In some earlier work the author has shown [17] how software may be brought closer in performance and reliability to hardware by various constraints on its execution paths. However, any such attempt in an IA would kill it in the sense that if the execution paths are frozen, the agent will no longer be intelligent.…”

“…Unlike hardware systems, where component failure rate can predict reliability, the reliability of software systems depends upon the probability of transition to an erroneous state during the execution of software. For a further discussion of this, the reader is referred to [17]. With further advances in the technology of artificial intelligence, and the emergence of large communicating systems that the Internet has made possible, there is yet another challenge to be faced by reliability engineers and researchers.…”