The Artificial Intelligence Research Center
Greeting from Director Japanese
In Artificial Intelligence (AI) research, maintaining the virtuous cycle between fundamental research efforts and their applications has become crucially and increasingly important. Diverse applications of AI, from financial engineering to biology and health have also meant that interdisciplinary cooperation among researchers and engineers from different fields and with different skills has become inevitable.
We have established the Artificial Intelligence Research Centre (AIRC) as a main venue, both in Japan and worldwide, which will (a) provide an environment in which leading and up-and-coming AI researchers of AI and its diverse related fields may carry out large-scale innovative work, with strong engineering support, and (b) facilitate close interaction between academia and industry.
In terms of research, we aspire to realize AI with a high affinity to human intelligence.
The recent development of Machine Learning-based AI has inspired high expectations but, at the same time, it has resulted in serious concerns or doubts that it may result in the creation of uncontrollable intelligence that is alien to human intelligence. Such doubts would lead to distrust of AI systems and restrict their usability in society.
Based on the huge memory and computational capacity of computers, machine learning is able to carry out inference over huge collections of data, which can reveal hidden regularities. Human intelligence cannot match AI in this respect. However, in some other respects, AI shows unexpected fragility or inability, which is never evident in human intelligence.
The different strengths of the two intelligences, i.e. human and artificial, makes it desirable to develop efficient means for them to cooperate in performing complex and challenging tasks. However, achieving this is dependent upon increasing the affinity of Machine-Learning based AI to human intelligence. This is essential to facilitate effective communication and feedback between humans and computers.
With this in mind, we perform research within the two following fields:
(1) Computing inspired by brain architecture and Neuro-computing: Development of flexible and robust computation methods for AI which emulate various aspects of computation in the human brain
(2) Data-Knowledge integration: Integration of Machine Learning inferences based on huge data with inferences based on knowledge and text which human society has accumulated
Each of these two ambitious research fields is broad and requires long-term research efforts with continuous innovations.
We pursue them through clearly defined application goals, and place an emphasis on close cooperation between academia and industry.