Visitor: Ryuichi Oka (Chief, Multi-Modal Functions/Information-Base Functions TRC Laboratory) Period: August 23-29, 1997 (7 days) Place: (attended conference): IJCAI-97 (Fifteenth International Joint Conference on Artificial Intelligence)
The Fifteenth International Joint Conference on Artificial Intelligence (IJCAI-97) was held in Nagoya. This is the second time for this Conference to be held in Japan, following the previous one in 1979. I participated to explain the demonstration (voice summary) at the AI academic showcase. I spent most of the time at the demonstration site, so I did not find enough time to hear the presentation of other papers. Therefore, this report only presents a single aspect of the Conference.
At this Conference, RoboCup, a group of soccer games played by robots gathered attention. Since the RoboCup site was next to the academic showcase, I could often watch the games. The RoboCup site and the AI academic showcase were popular with many visitors.
The AI-related researches have experienced many trends. Recent trends include the "Fifth Generation Computer Development", which was later followed by "Neural Fuzzy". The current trend may be "Complexity after Chaos".
In spite of these trends, I think there is a steady stream of expectations that AI research should pertain to practical use in society. In general, AI research, because of its difficulties, tends to give priority to topics which are academically interesting. Of these research topics, research on robots may be the best for incorporating reality into research. To create robots it is required to use visual, auditory, tactile, infrared, or supersonic sensors as input to implement moving and running by motion control. It is also important to implement the use of models and judgment functions for the advancement of robot development. Besides the achievement of these functions, the "gameness" of RoboCup can increase interest. These factors may have constituted the background for planning RoboCup. The actual robot games were still far from convincing the audiences of the "intelligence" of the robots, but I expect that achievements that were accomplished will advance further still for actual applications.
In the AI academic showcase, most demonstrations dealt with pattern information processing including man-machine interface, machine translation, voice translation, voice summary, and Jijo-2 robots (a knowledgeable robot). This may be because this conference was held in Japan, and "artificial intelligence" research flourishes in Japan; developments in such areas were presented in the academic showcase, the applications for each have gotten closer to real world applications. AI research has been out of vogue in the USA for a long time, partly due to a departure from handling real world information in a well-defined manner. On the contrary, AI research in Japan can be considered healthy.
Future implications of the "real world" are expected to expand AI research developments into the data communication world via networks. In the near future, therefore, the field in which the "artificial intelligence" operates will spread around the whole world over networks, and the stocked data will definitely be of a great variety, versatility, quantity.
RWC research topics considered to be guiding this type of research should be the most desirable. Regarding this, a recent article, "Artificial Intelligence: What Works and What Doesn't?" by Frederick Hayes-Roth in AI Magazine, Summer 1997, pp. 99-113, may be interesting. This article describes that the future AI systems should be those which evolve and learn, and whose functions and elements which vary in content and other aspects, should compliment one another to cooperate and compete. This is quite similar to the concept of "Information Integration" which RWC has been proposing for a long time. It is time for RWC to announce to the world the methods and algorithms based on the concept it was the first to propose on "Information Integration".