
Yoshua Bengio, a recognized pioneer in the field of artificial intelligence (AI), has recently expressed disappointment in the progression of Canada’s pending AI legislation. Known widely as one of the founding fathers of AI, and leader of Quebec’s esteemed AI research institute, Mila, Bengio has become increasingly concerned with the sluggish advance of Bill C-27.
The Digital Charter Implementation Act 2022, colloquially known as Bill C-27, also encompasses the Artificial Intelligence and Data Act. The Act aims to delineate clear boundaries for AI technology operations and passed its initial stages at the House of Commons in April. Despite its promising start, Bengio observes, the Bill remains stagnant at the committee review stage.
In a recent interview, Bengio expressed his anxieties about the delay in implementation, suggesting that parliament is not moving with the necessary urgency to legislate the rapidly expanding field of AI. The technology itself has progressed at an unanticipated pace, exemplified by the development of AI chatbot ChatGPT last November, a cutting-edge tool capable of intricate tasks and immersive human-like conversations.
Major tech companies, including Google, have entered the AI space, enhancing competition and innovation. The recent introduction of advanced AI developments incorporating voice and image capabilities heralds a new direction for this already revolutionary technology.
With an unpredictable trajectory leaning towards more influential AI systems, Bengio asserts both potential benefits and looming risks. The advancements in AI could offer profound opportunities, but similar to any potent tool, they also bear the risk of misuse.
The threats were underscored by Bengio and AI co-creator Geoffrey Hinton, who won the A.M. Turing Award in 2018 with Yann LeCun. Forewarning of the potential misuse of AI in areas such as job loss, bias, and even warfare, they encourage more comprehensive governance of AI risks.
In a document signed by approximately 20 AI authorities, the experts urged organizations and governments to dedicate one-third of their AI research and development resources towards the safe, ethical deployment of this technology. Additional recommendations included risk mitigation measures for AI technologies and corporate liability for potential harm.
Time is in essence, they argue, since AI has already surpassed human capabilities in areas including speed, knowledge absorption, and communication. Perceived risks include the exacerbation of global inequalities, surrogate warfare, manipulation, and increased surveillance.
A potential catalyst for the necessary change is Bill C-27. If approved, it will require companies to justify their reasoning for creating AI and mandate independent audits for compliance with AI laws. Yet, in lieu of legislation, a voluntary code has been proffered and accepted by a few corporations including BlackBerry Ltd., OpenText Corp., Cohere, Ada, Coveo, and Telus Corp.
Contrarily, Shopify Inc. declined to endorse the code, with founder Tobi Lutke advocating for innovation rather than regulation. When questioned about rallying oppositional voices such as Lutke, Bengio responded that swift legislation levels the playing field for all companies, ensuring collective compliance.
He comments that while self-regulation allows for accelerated resolution, it is merely a stopgap measure. Instead, swift, comprehensive legislation is a crucial requirement for the industry, ensuring that safety and ethics remain at the forefront of AI developments.