‘Godfather’ of AI says Canada must start regulating now

The government must regulate artificial intelligence (AI) quickly to protect Canadians from serious societal risks, said Yoshua Bengio, known as one of the ‘godfathers of AI.’ On Monday, the University of Montreal professor and scientific director of the Mila AI research institute told the House industry committee to adopt Bill C-27’s Artificial Intelligence and Data Act (AIDA) as soon as possible – despite its flaws – warning of serious risks if the technology continues to go unregulated. “An imperfect law whose regulation could be adapted later, is better than no law or postponing a law too much,” said Bengio. Because urgency is key and the regulations are expected to take at least two years to develop, he said some rules should take effect immediately after the law is passed. Powerful AI systems should be required to report to the government through a mandatory registry, providing information about their safety and security measures, proposed Bengio. He added that developers should be required to register their system and demonstrate its safety and security even before it is fully trained, in order to target “the risks that emerge earlier in an AI’s lifecycle, which the current law doesn't seem to do.” “The regulator will be able to use that information to form best-in-class requirements for future permits to continue developing and deploying these advanced systems,” said Bengio, suggesting that developers should be required to obtain a license to deploy such high-impact AI systems. “Businesses have to demonstrate that their product is not going to be harmful,” he stated, explaining that AI could be treated like other sectors that prioritize public safety, including medication, transportation, and food processing. Disinformation is one the main short-term risks Bengio sees AI posing to society, citing deepfakes and their potential to influence politics, but he noted that in a year or two, advanced AI systems could be used to develop cyberattacks and weapons. “Superhuman AI” that exceeds human intelligence “could be developed in the next two decades and even possibly in the next few years,” said Bengio. “We’re not ready, this prospect is extremely worrying.” The AI pioneer applauded the executive order that came out of the White House on Oct. 30 last year, which seeks to secure the development of powerful AI systems and reduce their risk on the national security of the United States. Bengio’s recommendations echoed some of what was stated in the order, including standards to identify and label synthetic content and requiring developers to share safety test results with the government. All witnesses at the committee meeting stated that making sure Canada’s AI regulations can align with standards emerging from around the world is important. McGill University’s Ignacio Cofone, Canada Research Chair in AI law and data governance, said improvements to AIDA’s definition of harms will help. Right now the bill is too focused on individual harms and does not recognize harms to groups or communities, said Cofone, pointing to examples of systemic discrimination and misrepresentation being built into certain AI models. Group harms like the impact of misinformation on democracy should also be recognized. “This fuller account of harms would put Canada up to international standards, such as the EU AI Act, which considers harms to public interest, to rights protected by EU law, to a plurality of persons, and to people in a vulnerable position,” he stated. Cofone also noted that it would align AIDA with existing ethics frameworks, like the Montreal Declaration on Responsible AI, the Toronto Declaration, and the Asilomar AI Principles. “AIDA is an accountability framework,” said Cofone, “and the effectiveness of any accountability framework depends on what it is that we hold entities accountable for.” Elissa Strome, executive director of the Pan-Canadian AI Strategy, agreed that quickly moving towards an adaptable regulatory framework is crucial, emphasizing the need to collaborate with global partners. “AI is not contained within borders,” said Strome, who pointed to the recent AI Safety Summit as a promising example of Canada collaborating with other countries on mitigating the harms of AI, even with those who might not be aligned on other issues, like China. – Reporting by Phalen Tynes-MacDonald at pcreed@thewirereport.ca and editing by Jenna Cocullo at jcocullo@thewirereport.ca

‘Godfather’ of AI says Canada must start regulating now

News | February 6, 2024

Unlock all the Canadian telecom, broadcasting and digital media news you need.

FREE TRIAL

SUBSCRIBE

Inside The Wire Report

Information

Hill Times Publishing

Get in Touch With Us

‘Godfather’ of AI says Canada must start regulating now

News | February 6, 2024

Unlock all the Canadian telecom, broadcasting and digital media news you need.

FREE TRIAL

SUBSCRIBE

Related Stories

Feds launch AI compute strategy

Canadian media companies sue OpenAI for copyright infringement, could cost $322 billion

OPINION: Government agencies using AI must implement governance programs, or risk losing trust

Champagne announces launch of Canadian AI Safety Institute

Reuse & Permissions

Inside The Wire Report

Information

Hill Times Publishing

Get in Touch With Us