If left without oversight, artificial intelligence poses a significant risk to humanity — a concerning message that Yoshua Bengio, a pioneer in the field, has been conveying for many years.
On Wednesday, alongside officials from the Canadian and German governments, a joint announcement was made regarding an investment exceeding a quarter billion dollars aimed at assisting the Canadian AI researcher in addressing this danger.
Bengio is working on developing a new type of AI framework, referred to as Scientist AI. This model is designed with safety as a priority; it aims to focus solely on objective truths and comprehend facts without having its own objectives — such as satisfying its human operators. Additionally, it could serve as a safety measure for other AI systems.
The federal government has allocated $150 million to LawZero, a non-profit organization founded by Bengio in Montreal last year, which aims to develop models like Scientist AI. This marks the first instance of federal financial support for the organization. Germany is contributing an extra 100 million euros.
Canada and Germany will make a combined investment of up to $300 million Cdn into a non-profit group founded by Canadian computer scientist Yoshua Bengio’s group LawZero. The goal is to help develop a new form of safe AI as part of a push toward Canada’s ‘sovereign AI capability,’ according to Minister of Artificial Intelligence Evan Solomon.
“I approach this with a sense of urgency because I want my children and grandchildren to inherit a future,” Bengio remarked during the All In conference, a significant event for AI leaders held in Montreal this week.
The apprehensions surrounding the risks associated with AI have escalated, particularly after a departing researcher from Anthropic — a leading AI firm — cautioned that AI could potentially eliminate humanity within the next ten years.
In a recent interview with Bloomberg, Prime Minister Mark Carney also advocated for the establishment of a global, non-industry organization to supervise and mitigate the risks of this technology.
-
Read more about LawZero: Can AI safeguard us against AI? One of its Canadian pioneers thinks so
-
Exclusive
Draft federal AI strategy aims to scale up adoption, offer literacy training by 2031
-
OpenAI flags 6 new examples of ‘concerning’ AI behaviour
Minister of Artificial Intelligence Evan Solomon stated that the funding for LawZero is part of Canada’s broader initiative for responsible AI development.
“We are convening today not merely in response to recent headlines, but following the philosophical stance we have adopted, which is to construct responsibly and safely,” he explained.
Safer by design?
Experts suggest that in many respects, the situation is already critical: it will be immensely difficult to address issues like AI generating false information oracting without explicit consent from humans.
“At present, there is no simple solution to this challenge. We possess extremely powerful, potentially transformative AI systems,” remarked Jaxson Khan, an authority on AI policy and integration at the University of Toronto’s Munk School of Global Affairs.
He noted that firms in the United States and China manage remarkably advanced AI systems, a trend he expects will persist.
Anthropic CEO Dario Amodei warned that AI development needs to slow down for safety measures to catch up. U.S. President Donald Trump dismissed the risk of artificial intelligence when questioned by reporters.
“At this moment, Canada feels a sense of limited control over AI,” he expressed.
However, partnering with Germany to invest in LawZero remains a significant step, according to Khan.
“It is essential to develop alternatives that are, ideally, designed with safety in mind,” he noted.
“I am interested to see how this unfolds, but I believe it is a valuable investment by the government and a beneficial collaboration with Germany.”
Domestic capacity and partnerships
Experts emphasize that collaborating with countries sharing similar objectives and values is another crucial element of the solution.
The announcement followed shortly after the merger agreement between two AI firms, Canada’s Cohere and Germany’s Aleph Alpha.
“This can be seen as a form of capacity building,” remarked Vass Bednar, managing director of the Canadian Shield Institute think tank, in reference to the merger.
-
AI ‘fearmongering’ could harm Canada’s rare opportunity, says top CEO
-
Analysis
One of the world’s biggest AI companies wants a deal with Canada. Is sovereignty the trade-off?
“We have the potential to catch up and create alternatives. We must also have faith in that,” she asserted.
“We cannot allow ourselves to succumb to the inevitability of technology or its adverse effects.”
Taking a multifaceted approach could lead to positive outcomes, according to Khan.
Bengio insists that surrendering is not an option.
“Reaching our objectives will not be easy,” he acknowledged. “However, if we can influence the world even slightly towards a better future for humanity, we must make the effort.”