Japan launches the Fifth Generation Computer Systems project

A national ten-year programme

In 1982 Japan's Ministry of International Trade and Industry (MITI) launched the Fifth Generation Computer Systems project, a ten-year national research initiative ^[1]. That same year the government established the Institute for New Generation Computer Technology (ICOT), created through joint investment with Japanese computer companies ^[1]. The Centre for Computing History records that in May 1982 ICOT published its plan as *Fifth Generation Computer Systems, ICOT Planning Report No. 1* ^[4].

The aim was not a faster version of the machines then on sale, but a different kind of computer: one built on massively parallel computing and logic programming, described in the project's goals as an "epoch-making computer" with supercomputer-like performance that would serve as a platform for artificial intelligence ^[1].

Logic programming and the LIPS target

Logic programming states a program as a collection of facts and rules rather than as a sequence of instructions. The machine answers a query by searching for a chain of rule applications that leads from the known facts to the query. Speed is therefore measured in logical inferences per second, or LIPS: how many such deduction steps the machine can take each second.

That measure set the project's headline target. The plan called for a prototype running at between 100 million and 1 billion LIPS, at a time when a typical workstation managed about 100,000 LIPS ^[1]. The work was scheduled in three stages: three years of initial research and development, four years building subsystems, and a final three years completing a working prototype ^[1].

Who carried it out

Feigenbaum and Shrobe, in a survey published by Stanford University, write that the project was planned for ten years and was carried out by a central research institute, ICOT, with personnel from its member-owners: the Japanese computer manufacturers and other firms in the electronics industry ^[3].

The choice of logic programming shaped the staffing and the software agenda as much as the hardware. The machine was meant to run programs written as rules, so the language and the programming methods were part of the research problem rather than something inherited from existing practice.

The machines

Five working Parallel Inference Machines were produced: PIM/m, PIM/p, PIM/i, PIM/k and PIM/c ^[1]. The project also built sequential inference machines (PSI) and parallel inference machines (PIM), and the performance of the combined hardware and software was measured within the planned range of 150 million logical inferences per second ^[3].

On the software side, ICOT chose to concentrate on language issues and programming methods for logic programming. A system for logic programming on parallel machines, called KL1, was constructed, but applications were carried out in demonstration form only ^[3].

Outcome

The project ultimately ended in commercial failure, although its ambitious goals contributed significantly to concurrent logic programming ^[1]. Its highly parallel architecture was eventually surpassed in speed by less specialised hardware such as Sun workstations and Intel x86 machines ^[1]. More broadly, existing CPU performance quickly overcame the barriers experts had anticipated in the 1980s, which reduced the value of parallel computing ^[1].

The Centre for Computing History states that on 1 June 1992 Japan formally wound down the Fifth Generation project, and that despite its grand aims it had little impact on the global computer market ^[4]. Wikipedia gives the project's span as 1982 to 1994, with a total spend of a little less than ¥57 billion, about US$320 million ^[1].

Reactions elsewhere

The project was followed closely outside Japan. Wikipedia notes that Cyc, the long-running common-sense knowledge base project started at MCC, was begun partly to counter the Japanese fifth-generation effort ^[2].