The Quebec Innovation Council published Monday morning a dozen recommendations to help the Quebec government properly regulate artificial intelligence (AI). If they are inspired by bills already existing elsewhere in the world, they also highlight the importance for the province of stimulating the creation of a uniquely Quebec AI.
“By creating the structure behind an expanded model of Quebec language, we will be able to bring our historical archives, our literature, our culture and our knowledge back to digital life,” explains the Duty Quebec’s chief innovator and general director of the Quebec Innovation Council (CIQ), Luc Sirois.
An expanded language model, called LLM in English, is the engine behind AI applications like OpenAI’s ChatGPT, which is called GPT. These models are based on the accumulation and analysis of a massive quantity of data drawn from the digital universe. However, the most popular LLMs currently were created by researchers in the United States, which creates a bias to the advantage of American culture in the way information is interpreted by these AIs.
“Having a local research and development capacity, so as not to have to depend on others, can accelerate the creation of a sovereign AI here,” explains Luc Sirois. “It will also benefit local businesses. »
It’s up to the government
Mr. Sirois presented his report Monday morning in the company of the Minister of Economy and Innovation and Minister responsible for Regional Economic Development, Pierre Fitzgibbon, as well as the Minister of Cybersecurity and Digital Technology, Éric Caire.
In government, there seems to be agreement on the economic potential of technology. We also understand that AI comes with its own risks. The Ministry of Cybersecurity and Digital Affairs has also reorganized its organization chart in recent weeks to make more room for a unit that will focus specifically on artificial intelligence.
Sarah Gagnon-Turcotte, former leader of the CIQ and former right-hand man of Luc Sirois, has also joined the ministry. The latter is delighted to see her present in the heat of the action, so to speak. “She embodies our recommendations. I am reassured that she is there, she participated in all the debates and consulted all the experts. »
Indeed, despite its numerous recommendations, the Innovation Council is well aware of one thing: legislating too much could pose almost as many risks for Quebec as not legislating at all. This is why he suggests an approach where the responsibility of creators of AI applications would be modulated according to the level of risk that their technology entails.
This approach is seen in the industry as the best balance to encourage innovation while reducing the risks of slippage and the use of technology for malicious purposes.
High-level recommendations
The twelve major recommendations of the Quebec Innovation Council are classified into five areas. The organization wanted to supervise the development and use of AI, anticipate its future applications, train and promote local researchers and businesses and position Quebec at the forefront of the industry.
One of the challenges of the exercise was not to overdo it. The federal government also has a bill to regulate artificial intelligence, which is currently under study and which could be adopted in the coming months. Ottawa has meanwhile published a voluntary code of conduct to prepare the ground. The European Union also has its legislative framework, and the President of the United States presented a decree just before Christmas to also avoid slippages.
In short, this suddenly creates a lot of constraints for the technology sector. This is, among other things, why Quebec should settle for a high-level framework law, then delegate responsibility for regulations to an independent authority, argues the CIQ.
This is also why its recommendations speak of the importance of ensuring the development of “responsible and sustainable” AI, but avoid specifying how to get there. For example, we could have required the AIs present in Quebec to “sign” the content they generate so that the public can quickly distinguish it from more authentic content.
Adapting professional training to prepare workers for the emergence of AI is also among the CIQ’s other suggestions, which goes further: it invites the entire education system to strengthen digital literacy and understanding of technologies by students.
Finally, the Council proposes that the government make a budget available to public and government organizations so that they can in turn guide the use of AI in their operations. Public sector employees should also be better prepared for the rise of new technologies, whether to better use them at work or to avoid the traps set by more malicious users of these technologies.
Indeed, generative AI makes it possible to generate text as well as voices, photos and fake videos sophisticated enough to lure people who are not paying particular attention.