State of the art and a glance at the future
Many criticisms have been made of current artificial intelligence systems: it is obvious that they are poor and disappointing compared with initial expectations of artificial intelligence. In fact, we have not seen giant leaps forward; the most demanding problems, e.g. learning and the representation of common sense, although they have been partially dealt with and solved, are far from being solved completely. Nevertheless, expert systems are very common today from a practical and commercial point of view, even if acquiring knowledge of them is without doubt the bottleneck which prevents them from being used more. Fully extracting knowledge from an expert and succeeding in formalizing it in a knowledge base is particularly complex. Furthermore, these systems cost a lot to maintain and update. However, neural networks are applied nowadays with success, although they are often limited to solving problems seen as low-level, such as perception and recognition.
As far as the future is concerned, the current technological revolution that is providing today’s society with information allows us to access an enormous mass of material, but this information must be managed and interpreted correctly. It therefore seems right to try and not only strengthen but also and especially revolutionise the instruments that extract and analyse information, in order to maximise the potential of this richness of knowledge. It is therefore essential to use the previously mentioned methods to extract knowledge, using techniques of symbolic learning and neural networks.
There is a big push at the moment in favour of the integration of artificial intelligence systems, in particular expert systems, with the rest of the information engineering world.
One important phenomenon is the gradual extinction of the open system, taken to mean a stand-alone application, in favour of an integrated system. There is now the tendency to create modules that produce "intelligent" tasks, strictly integrated in software applications and in general information systems. The idea is therefore to build "intelligent agents" capable of deductive and inductive reasoning, prepared for particular tasks and able to coordinate with other agents in a distributed architecture in order to achieve a single aim together. In these fields, artificial intelligence is already meeting with and will continue to meet with considerable success.
information on the activities of the artificial intelligence research group at
the Department of Electronics, IT Techology and Systems Analysis (DEIS), Faculty
of Engineering, University of Bologna, see the following link: