One of the biggest disappointments for the computer industty 1980s was the failure of knowledge-based systems to occupy more than a marginal role in mainstream data processing, but Mountain View, California-based Market Intelligence Research Corp reckons that their most promising successors, neural networks and fuzzy logic are about to find their place in the high-tech sun. The market for technologies that seek to simulate human reasoning will grow worldwide by more than 30 times over the next seven years according to the researchers. They put the 1991 market at just over $300m and are shooting for nearly $10,000m by 1998, expanding at a compound annual revenue growth rate of nearly 65%. According to the report Imitating Human Reasoning: the Viability and Commercialisation of Neural Networks and Fuzzy Logic, recent growth rates of over 50% will accelerate into the three-digit range in the coming years, with the market nearly doubling every year until 1996. Rather surprisingly, the report suggests that the two markets are now of roughly equal size the perception is that while neural networks are still mainly laboratory toys, fuzzy logic, thanks to its widespread use in consumer appliances from Japanese manufacturers, is now in the main stream – the downside presumably being that unit values of fuzzy technology in washing machines and so forth are low. At all events, the researchers reckon that fuzzy logic will grow more rapidly until mid-decade, when neural networks will recapture first place. Neural networks have a projected compound annual growth rate of 46% between now and 1998, while fuzzy logic is projected to grow at 76% compound over the same period.
Synergistic rather than competitive
The two technologies are forecast to be primarily synergistic rather than competitive, and to coexist on a common integrated system in the future, combined with related technologies like expert systems. The researchers reckon that while Japan holds the ring in fuzzy logic, the US has led in neural network development, although new interest by Japanese companies should promote competition while the US switches on to fuzzy logic, witness Motorola Inc buying a licence to a set of fuzzy algorithms that it wants to embed in silicon. While Europe has lagged in both arenas, major European competitors, such as Philips Electronics NV, Siemens AG and Thomson-CSF SA have started to move into both. Just as most knowledge-based systems are written in languages such as C and C++ rather than Lisp or Prolog these days, standard software and integrated circuits will displace engineering development tools and customer applications as the dominant product segments, although the latter make up 90% of the combined markets at present, the researchers say. Financial and industrial applications for neural networks will expand their share of the market worldwide at the expense of the defence segment, while medical applications will also expand substantially. The researchers say the report suggests that entry barriers are not high in the emerging markets and competition is not extreme – needless to say the main obstacle is the availability of highly qualified personnel. The researchers do not quite go overboard in their forecasts, saying that despite rapid advances, neural network and fuzzy logic technology will not challenge RISC or parallel processing for dominance as a microprocessor architecture before the turn of the century. The report was not received for review and the company did not say how much it charged for it.
