Cybernetics has made a very modest progress resolving the tasks it was intended to by its creators. Even today, after 50 years of the worldwide research efforts there is no existing automatic system able to control processes as complex as those that are easily controlled by the biological systems. We think that cybernetics was unable to hold its promises just because the central feature of biological systems — ability to learn from experience was never sufficiently developed in artificial systems.

Even the best software today doesn't utilize previous experience making the same mistakes or reanalyzing the same data every time it starts. The burden of input and conditions generalization lies on the developers' shoulders at the software design and development time. If the state of the world moves out of the initial assumptions the software goes out of synch. The everlasting Year 2000 problem is a good example of such problem. Less general, but sometimes even more painful examples can give you any experienced IT manager. They all know how much work end expense it takes to adjust a large interconnected software system to new requirements or even just to increasing volume of processing. Here again an adaptive, self-learning, live system, optimizing its behavior dynamically at run-time should replace today's code — deaf to its own results, not optimal by definition.