Nobody yet has a hard-and-fast definition of a fifth generation computer yet, but that has not deterred Professor Takeshi Yamakawa of Kumamoto University from building a prototype of what he describes as a sixth generation computer, Kyodo news agency reports from Tokyo. Key feature of the machine, developed in conjunction with Omron Tateisi Electronics Co, is that it applies fuzzy logic, and is designed to process imprecise natural language input. We have just mastered the concept of LIPS – logical inferences per second, now come FIPS, fuzzy inferences per second. The prototype is rated at 10 MegaFIPS. Fuzzy logic represents data in terms of grades of membership going from 0.0 – no connection, to 1.0 – the same. As a result, a fuzzy computer is designed to handle concepts like more and somewhat. The prototype consists of a fuzzy memory, a set of inference engines, a max block, a defuzzifier, and a control unit. It converts linguistic information into an analogue voltage ranging from zero to five volts on the data bus, but the information is stored in binary form using dynamic RAMs to get round the problem that external interference can alter the voltages, thereby destroying the data. The partners have designed emitter-coupled fuzzy logic gates which respond in under 100pS, and have also designed truncation gates, which truncate the membership function and allow the inference to carry out minima-maxima operations.