July 23rd, 2024

AI pioneer Yoshua Bengio wishes Canada’s AI legislation was further along by now

By Tara Deschamps, The Canadian Press on October 24, 2023.

Several Canadian pioneers in artificial intelligence have signed a new document urging governments to better manage the technology's risks as its capabilities continue to advance rapidly. Yoshua Bengio, founder and scientific director, Mila-Quebec AI Institute, discusses artificial intelligence, democracy and the future of civilization at the C2MTL conference, in Montreal, Wednesday May 24, 2023. THE CANADIAN PRESS/Christinne Muschi

TORONTO – Artificial intelligence icon Yoshua Bengio wishes Canadian legislation on theincreasingly mighty technology he pioneered would be further along by now.

Bengio, who is best known for his status as one of the godfathers of AI but also runs Quebec-based AI research institute Mila, expressed his disappointment with the pace of Bill C-27 in a Tuesday interview.

The bill, called the Digital Charter Implementation Act 2022, includes an Artificial Intelligence and Data Act, which would place guardrails around the use of AI. It passed its first and second readings in the House of Commons in April but has yet to progress beyond the committee review stage.

“I was hoping six months ago that the proposed law would come into effect this year,” Bengio said.

“I don’t know if this is going to happen. There’s not much time left. Parliament is slow.”

Bengio would much prefer governments move forward with urgency when addressing AI because the technology has advanced at a pace that even he could not have predicted since ChatGPT, a generative AI chatbot capable of humanlike conversations and tasks, was launched last November.

The chatbot, developed by San Francisco-based OpenAI, kick-started an AI race with top tech names including Google and its rival product Bard and a slew of startups innovating in the space. A new iteration of the technology with voice and image capabilities was released this month.

“We are on a trajectory that is uncertain but clearly going in a direction of more and more powerful AI systems that may eventually reach human capacities in many domains,” Bengio said.

“That is both exciting in terms of the good things that we could do but also dangerous because any powerful tool could be misused.”

His concern is so heightened that he and fellow AI godfather Geoffrey Hinton – they won the A.M. Turing Award with Yann LeCun in 2018 – have been sounding alarm bells about the potential for the technology to be used in dangerous ways for much of the year. Hinton has said the technology’s risks include bias and discrimination, joblessness, echo chambers, fake news and battle robots.

On Tuesday, they doubled down on their warnings, releasing a new document written with roughly 20 other AI experts that urges governments to better manage AI’s risks.

To keep up with the pace of AI innovations and mitigate risks, the signatories are calling on governments and companies to allocate one-third of their AI research and development funding to the safe and ethical use of the technology.

They also want creators of AI systems to adopt safety measures that can counter their products’ capabilities and companies who make these technologies to be held liable for any harms they may cause.

The time to act is now, they say, because AI has already surpassed some human abilities since these systems can act faster, absorb more knowledge and communicate faster than humans.

Further advances mean AI could exacerbate global inequities, facilitate automated warfare, orchestrate mass manipulation and pervasively surveil people.

So far, the technology has left humanity in a “chaotic, uncontrolled race,” said Bengio.

“The governments have no handle on (AI) and general public has no handle on it,” he said. “This needs to change quickly.”

Some of the change could come from Bill C-27, which will force firms to document their rationale for developing AI and set up a commissioner to independently audit companies for compliance with AI laws.

In the absence of legislation, the government is using a voluntary code signed by BlackBerry Ltd., OpenText Corp., Cohere, Ada, Coveo and Telus Corp.

The companies have agreed to test AI systems in advance of deploying them to uncover potential risks and biases and then promise to implement measures to mitigate the dangers and maintain a database of any incidents.

E-commerce giant Shopify Inc. refused to sign the code, with founder Tobi Lütke saying, “let other countries regulate while we take the more courageous path and say “˜come build here.'”

Asked about how to get detractors like Lütke on board with such guardrails, Bengio said, “you can’t and that’s why we need to hurry and to make it a level playing field for all companies.”

“It wouldn’t be fair if the good guys do the extra work to protect the public and some just take advantage of the situation to carve a larger part of their market,” he said.

“Self-regulation is a temporary thing that can move faster “¦ than the pace at which democracies are able to come up with rules as quickly as possible, but it has to become something entrenched for all companies.”

This report by The Canadian Press was first published Oct. 24, 2023.

Share this story:


Comments are closed.