Search results

It is often argued that anything observable may be simulated on a computer. Using this as a basis, workers in artificial intelligence (AI) often go forward to maintain that machines can be made intelligent by machine simulation of human intelligence processes. There are two difficulties with this concept. The first difficulty lies in the knowledge of human intelligence processes that we have presently obtained and may possibly obtain in the near future. A more basic question is of the sufficiency of the concept itself. Simulation in itself is not sufficient to produce intelligent action where perhaps modelling might be. There are fundamental difficulties in the problem of establishing an adequate mapping function. It is held that there is insufficient correspondence between human and machine intelligence processes to allow human intelligence modelling on existing digital computers.

Knowledge‐based systems have been successfully utilised in the develop‐ment of complex systems. In many cases, these systems have emphasised the need for techniques to integrate knowledge‐based processing with methods for managing both large amounts of data and knowledge. However, many potential applications for expert systems are precluded by limitations in the ability of conventional expert system technology to function in conjunction with data systems without manual intervention. The author focuses on the integration of knowledge‐bases and databases with the capability to: store and context select between parallel, competing expert system rule structures; cascade variable rule structures; allow an expert system to be interrupted and to be subsequently restarted by storing the state of the inference engine; handle simple data storage and retrieval.