Last week, the Canadian government offered its support for the Seoul Declaration for “safe, innovative and inclusive” artificial intelligence. This support will be wishful thinking as long as Canada’s artificial intelligence (AI) strategy is inspired by the speculative model of Silicon Valley and serves foreign interests that have nothing to do with the well-being of the Canadian population. .
The massive investment of two billion dollars to create public access to efficient computing infrastructure could, in return, represent a golden opportunity to harness AI for the benefit of the common good.
Currently, access to high-level computing power is limited because only very large companies can set up adequate infrastructure. According to information contained in the latest Freeland budget, the government recently signed a letter of intent with the company Nvidia. The latter manufactures electronic chips that are excessively expensive, but extremely efficient for data analysis. If the government wants to give a boost to AI companies and researchers, it is because it believes in significant economic benefits for Canada. Yet promises of gains have been undermined by research that demonstrates the pitfalls of the current development strategy.
In 2019, the Institute for Socioeconomic Research and Information (IRIS) published research that showed economic expectations for AI were inflated amid increased corporate innovation concentration. This meant that Quebec and Canadian companies had little chance of generating social benefits since they were doomed to be bought by Silicon Valley giants or to go bankrupt. Since then, Element AI and Imagia, two companies perceived as the flagships of the future of AI in Quebec, have respectively followed these avenues.
Other work has also shown that those who mainly benefited from public investments were a very small number of companies and researchers in the sector, whose intellectual property Canada was unable to retain. Furthermore, private investment in research and development continues to decline, while the results of public science are increasingly privatized.
As for the effects on productivity, these have yet to be demonstrated, especially in the public service. In the health sector, IRIS showed that the dominance of the pharmaceutical and digital industries was preventing an overall increase in efficiency in the public system. The technologies developed serve hyper-specialized medicine which almost entirely neglects prevention. Interventions on the environmental or social determinants of health are left aside, to the benefit of industries that profit from the disease.
Finally, let us remember that this infrastructure will be extremely energy-intensive and that the increase in computing power weighs increasingly heavily on Canadian energy demand. It is difficult to believe, in these circumstances, that these computing infrastructures will unequivocally serve broader and lasting interests.
Should we therefore refuse public investments in computing infrastructure? Not necessarily. In fact, the idea of such an infrastructure presents an opportunity to create a new model for AI development. Rather than producing technologies that fit into the deadly economic logic of the powerful Silicon Valley giants, innovation policies could stimulate the creation of technologies that have a real impact on the work and lives of the Canadian population. Whether by imposing the creation of freely accessible innovations or by favoring projects that engage in co-production approaches, a public computing infrastructure must avoid the current pitfalls of AI development policies. Only with such ambitions will citizens be able to believe that they will change their ways.
What is the point of implementing a strategy for sovereign computing power without considering technological sovereignty?