During the 1980s, Japan embarked on an ambitious project called the Fifth Generation Computer System (FGCS) to achieve the ‘AI dream ‘ while the United States was experiencing an AI Winter. Japan had lagged behind the US in technology for years and played follow-the-leader. In 1978, Japan’s Ministry of International Trade and Industry (MITI) commissioned a study to predict the future of computers. Three years later, they attempted to build fifth-generation computers that project heads described as a significant leap in computer technology, giving Japan a technological advantage for years. These high-power logic machines specialized in logic programming rather than standard microprocessors, hoping to catalyze the world of information processing and realizing artificial intelligence.
After ten years of research and over a billion dollars in funding, we still require advanced computers. Most of the studies did not push the boundaries of technology. Where it did, it was only in areas specific to the scale and scope of FGCS applications. Nevertheless, the story of the FGCS project is fascinating and worth examining. The grand vision of intelligent machines and the epic failure of the project are both deserving of investigation.
Motivations and influences
Japan’s motivation for building FGCS machines was to advance technology and take the lead in the computer industry, similar to their success in the automotive industry. A study determined that knowledge processing would be the future of computers, and Japan could benefit financially by developing specialized machines for this task. The director and visionary of the FGCS project, Kazuhiro Fuchi, believed that knowledge information processing was the way forward for information processing technology.
The project aimed to create KIPS or Knowledge Information Processing Systems, which would not rely on fancy algorithms or learning but instead iterate over massive knowledge bases to infer information. These systems would use hardware-based syllogism engines to deduce information. Logic programming has been adequate but limited to academic environments. The FGCS project hoped to create specialized hardware that could process massive databases instantaneously.
The project aimed to develop machines capable of making a million logical inferences per second, which could translate natural speech, prove theorems, and perform other intelligent activities. MITI formed ICOT to achieve this goal, and the project took ten years to complete. The funding for the project had salaries for employees from Japanese technical firms who were expected to invest in the FGCS project after it demonstrated results. The project aimed to create a critical mass of knowledge that could be contained in a database file system and processed instantaneously.
Progress made on FGCS
Successes
It’s clear that a lot of resources and effort were invested in the FGCS project, and it’s not surprising that there was much pressure for it to succeed. Fortunately, the project produced impressive results, with machines capable of making many logical inferences per second. Seeing how this technology grows in the coming years will be unique.
Failures
Perhaps the failure of the FGCS project was best outlined in a statement at the meeting preceding the completion of the intermediary stage of the FGCS project: “Most of the applications presented at the conference were interesting because they were ‘X done in logics programming’ – not because they were ‘X done better than before.’ The course hopes the final computer will be fast enough to run infeasible programs on regular computers.
The main issue was the failure to realize ‘intelligent’ software. The IOTC hired experimenters to tackle intricate problems, including natural language processing, automated theorem establishing, and a program for playing the game of Go. Unfortunately, these fields did not see notable advancements due to upgraded hardware. Despite possessing ultra-powerful logic machines, I was teaching software to be intelligently proved to be an intimidating task.
Like Type-A vs. Type-B chess programs, you can convert computational power into increased perceived intelligence in some situations. However, the bottleneck is typically understanding how to teach the computer to think – by abstracting the problem space. Another problem with the project was the assumption that parallel logic chips would be required to perform advanced computations. Microprocessor technology advanced steadily during the ten years of the FGCS project. While FGCS hardware was proficient in pure logic programming, commodity hardware proved a worthy competitor, particularly compared to low-end single-processor KIPS systems. Processor KIPS systems.
Lessons learned
Despite the high hopes, the FGCS needed to realize the enormous advances expected. Some issues, such as an emphasis on parallelization, were ahead of their time, while others, like massive knowledge databases and pure logic-based programming, didn’t pan out. It is easy to disregard the FGCS as a failure in history, but we can learn some powerful lessons about government-mandated research and a push toward intelligent systems.
Government-sponsored computer research has worked well for the United States; most of the history of early computers is owed directly to US military spending. Japan’s approach to the FGCS project distinguished itself from other US-based research endeavours like DARPA by dedicating all its resources to creating intelligent machines. While a focused effort in a single area/technology can lead to more enormous rewards, it is difficult to predict whether that technology will succeed.
With Japan’s goal of becoming a world leader in computer technology, it would only be possible to ignite a revolution by taking risks. So it is difficult to notice this as a mistake. If the project was successful and did find ways to increase software intelligence through customized hardware dramatically, then the world of computing as we know it would have been forever changed. Just as US-based experimenters in the 50s and 60s theorized that appliances would be qualified to pass a Turing Test ‘soon,’ the problem at the root was underestimating the difficulty of artificial intelligence.
The most important part of the FGCS project was the gamble created on intelligent computers. There was never a debate or hesitation on whether a machine could be brilliant, just a bold attempt to realize that vision. Perhaps the lesson to learn is that such a view is unsupported, and the defeat of legions of Japanese researchers should provide doubt about the usefulness of AI. More likely, the bottleneck in artificial intelligence development is software rather than hardware. Maybe another FGCS task is even necessary for the development of true AI, but probably only after we create intelligent machines that are required to be faster for functional use.
Conclusion
The above sections have explored aspects of AI history that share common themes. The field is pulled in two directions, one towards practical applications in the short term and the other towards more significant issues that challenge the very definition of intelligence. Considerable progress has been made in AI over the last 50 years, leading to the developing a well-defined field that has solved various problems, such as adaptive spam blocking, image/voice recognition, high-performance searching, and more. However, despite this progress, the goals of pioneers such as Turing and McCarthy still seem more distant than ever.
AI has faced challenges in solving complex problems, resulting in slow progress due to reduced funding, lack of interest from researchers, and the realization that issues become increasingly difficult to tackle with greater understanding. These challenges have contributed to the possibility of an AI winter scenario, where resources are diverted to more practical applications. Despite 50 years of work, no computer has successfully passed the Turing test, there has yet to be widespread replacement of experts with software systems, and humans still outperform computers in certain simple yet strategic games like Go.
This doesn’t mean that we give up thinking and trying, but just that we refine our approach. Over the years, we have learned that more than a massive knowledge base and a million logical inferences are needed in a second. The complex problem in AI is finding a way to teach a machine to think but to articulate ‘thought’ in a way current computers can understand; we must first understand thinking and intelligence ourselves. Until then, we will build chess programs that depend on brute force, expert systems that manage the obvious, and talk with programs that aren’t attractive to talk about.
