mack (02/24/83)
Towards a National Policy for Artificial Intelligence: Notes for a Discussion of AI in Canada Alan Mackworth and Ray Reiter Department of Computer Science University of British Columbia Vancouver, B.C. V6T 1W5 1 The Need for a Policy 1.1 AI will be a central component in future computing environments. Supporting evidence: (i) The Japanese 5th generation project. (ii) US industrial AI research labs. (Fairchild, TI, Xerox, SRI, Schlumberger etc.) (iii) US academic AI research labs. (MIT, Carnegie, Stanford etc.) 1.2 Canada should not become an AI (and hence high tech- nology) 3rd world nation. Apart from the obvious shift in the economy from resource exploitation and manufacturing to knowledge-based industries, there are other reasons unique to Canada that dictate substantial, focussed support for AI. The most important of these are: - The emphasis is on software not hardware. Software is increasingly dominating the cost of computer sys- tems. - AI development will help the resource industries tremendously through remote sensing and Prospector- like expert systems. - Robotics and advanced automation require high-level reasoning, planning and vision systems. - Our bilingual environment dictates work on language understanding and translation. 1.3 Canada has no privately funded AI research labs of any consequence. Its AI expertise is concentrated in academia, primarily in computer science departments. These departments are small, overwhelmed with undergra- duate students and seriously underbudgeted. The US offers far better AI research environments which are increasingly drawing away some of our best AI research- ers and graduate students. In particular, these stu- dents, on graduation, tend to remain in the US where they are attracted by high paying industrial positions - 2 - and attractive research environments. This trend will accelerate if nothing is done to encourage Canadians to remain or return here. 1.4 The current situation in Canada is a result of under- funding of Computer Science by NSERC and a neglect of AI by the Canadian academic Computer Science community. The three leading academic computer science environments in the US (Stanford, MIT and CMU) and many others (Rochester, Yale, Rutgers, Columbia, Amherst, Maryland, UC Irvine, UC San Diego, Texas, etc.) have each developed AI as a central area of competence. In Canada, only one or two departments have pursued that strategy. In the US, ARPA has provided massive support for AI, and NSF created the Intelligent Systems Program for AI and an Experimental Computer Science Program. 2 Background 2.1 Basic AI research in Canada is, and will remain for some time, centred in the universities. 2.2 Canada must remain abreast of current and future developments in the field. At least initially, this responsibility rests with its academic AI researchers. 2.3 The proper national and economic role for Canada's academic AI research community is to train future gen- erations of AI professionals who are, in turn, applying their expertise in Canadian industrial, entrepreneurial, and academic settings. 2.4 There are already centres of excellence in specific areas of AI across the country. Their development should be encouraged and accelerated with common infras- tructure support available over the new computer net- works. 2.5 The application areas currently of most promise for the Canadian economy are: (i) Office Automation and Databases (ii) Robotics (iii) Remote Sensing and Resource Management (iv) Natural Language Translation and Interfaces (v) Education (vi) Medicine There are existing Canadian research groups in each of these areas that must be strengthened. We emphasize, however, that in such a young field the science and engineering (theory and application) must develop together; for example, concentrating exclusively on - 3 - "knowledge engineering" or expert systems for specific applications would be a short range sterile strategy. We therefore believe that a national policy for AI is necessary and that it should focus on strengthening the academic research environment. This is where Canada's current expertise is concentrated. This is where future expertise will be trained. 3 Proposals A national industrial strategy for Canada should recog- nize the central role of artificial intelligence and com- puter science. The Science Council's earlier reports on high technology have already changed the perceptions of pol- icy makers and planners. A planning document for AI development would perform a similar function, suggesting roles for government, industry and academic groups. In par- ticular, NSERC should take an activist position. We propose that NSERC declare AI a high priority field. This should minimally entail: 3.1 Increased funding for research and graduate student sup- port. 3.2 Provisions for partial release time from administration and undergraduate teaching duties for selected research- ers, a common practise in US universities. This is especially important now in computer science departments where everyone is seriously overworked. 3.3 The addition of Artificial Intelligence to the set of areas funded under the Strategic Grants Programme. 3.4 The provision of state-of-the-art computing facilities for AI researchers. These are now cheaper and more powerful than previously but a capital investment of $50K-$150K per researcher is necessary (the level depends on the area of research). This should be recog- nized in the Strategic Grants program for AI. One or two of the existing AI centres should be funded for AI software development: to maintain, develop and import/export useful software tools. They should be supplied with the necessary infrastructure support to operate on a long-term basis. 4 Conclusion 9 9 - 4 - The implementation of these proposals is essential for the creation of an environment that would encourage our ablest researchers to remain in Canada, and induce some of our expatriates to return. It will also allow researchers to focus on graduate education, strengthening graduate pro- grams in AI, and therefore attracting graduate students to Canadian universities. Such students are then more likely to remain in Canada upon graduation. An important conse- quence of the existence of a body of experts representing a broad spectrum of AI is their availability as independent consultants to industry, entrepreneurs and government policy makers. We believe the benefit/cost ratio of the proposals made here to be among the highest that the government could achieve in any scientific or engineering field. The economic, social and intellectual benefits are greater than those of the classical sciences. The cost of not making this investment in the future would be enormous. 9 9