Softer than soft computing

In hard computing for engineering applications, we use explicit models derived from physical principle, and implement them on a computer as purely syntactic Turing-equivalent structures. When such a direct attack is not feasible, we resort to soft computing, using the techniques arising from artificial intelligence to ferret out the secrets of a process based on implicit models derived from observed data. Again, we implement them on a computer as purely syntactic Turing-equivalent structures. As the interest of the engineers moves toward problems in biomedical engineering and human-machine interaction, it is apparent that there are problems intractable even by the methods of soft computing. Processes of life and mind include internal semantics including inherent semantic ambiguity that are indispensable to their operation, but these semantics are totally missed by the purely syntactical strategies of both hard and soft computing. For engineers to make responsible decisions about systems that involve naturally occurring processes of life and mind, a new modeling strategy is required. It needs semantics models that can account for internal ambiguity and has so high a degree of flexibility that we may think of it as softer that "soft computing".