The United Nations’ top technology official fears that corporate interests are undermining efforts to harness artificial intelligence (AI), exacerbating social divisions and encroaching on human rights.
Countries may feel pressure to respond to business demands for greater leeway rather than curbing industry excesses, Amandeep Gill said in an interview on the eve of a global conference on ‘IA in Montreal which starts Wednesday.
“Frankly, I’m quite worried,” he said.
Human rights and democratic values are at stake, said Secretary-General’s Special Envoy for Technology Gill
Researchers and policy leaders have highlighted concerns ranging from subjective data sets and worsening global inequality to existential threats from large-scale cyberattacks and AI-developed biological weapons.
Yoshua Bengio, pioneer of artificial intelligence and founder of the Mila Institute, has sounded the alarm about immediate dangers such as “counterfeiting humans” using AI-driven robots.
Academics, business leaders and policy makers are in Montreal for the three-day gathering organized by Mila.
The consolidation of power, prejudice and privacy are three of its central themes.
“As there is potentially a greater concentration of wealth and technological power in a few companies, that has implications for social equity, for our social contract,” Mr. Gill said.
The desire to get a head start in the global technology race could conflict with the need to reduce risks from rapid advances in AI through laws and regulations, he argued. Mr. Gill also pointed out that oligopolies or the concentration of AI in a handful of countries would disadvantage small businesses and developing countries.
Increased polarization within and between states is another possible outcome of AI slip-ups.
“If AI is exploited for hyperfaking, misinformation and disinformation on a large scale, it could undermine the legitimacy of political processes in our society,” he added.
AI bias
Inherent bias is also a recurring problem in the industry, as algorithms that rely on reams of data sometimes amplify existing biases rather than eliminate them. “This could continue and even worse,” he said. Whether decisions are made regarding housing, parole, the justice system or the allocation of social benefits.”
Discriminatory results are not limited to the national level, Gill said, adding that information fed into machine learning models largely comes from North America or Western Europe.
“This means that a vast majority of the world’s cultures, languages and contexts are not adequately reflected in these datasets. »
At the same time, the power to track the online activity of citizens and social media users risks leading to violations of the right to privacy.
Despite the urgency of controlling cutting-edge AI, Gill said the United Nations must take a “modest” approach to setting rules, to encourage as many states as possible to sign up.
Last month, a UN advisory body released a preliminary report setting out the guiding principles for a framework on AI governance, emphasizing that no country should be “left behind” as the pace of innovation approaches the speed of light.
A bill under study in Canada
In Canada, the federal government introduced a bill placing guardrails around the use of AI in June 2022, but it has been languishing in committee for almost ten months.
Executives from major tech companies said last week that the Artificial Intelligence and Data Act is currently too vague, arguing that it fails to distinguish between high- and low-risk AI systems.
The Liberals said they would change the law to introduce new rules, including requiring companies responsible for generative AI systems — the algorithmic engine behind chatbots such as OpenAI’s ChatGPT, which can provide any anything from math problems to marriage advice — to taking steps to ensure their content is identifiable as being created by AI.
The Bill still aims for a more general and principles-based approach to AI governance, which allows for some agility in the context of constantly evolving technology, leaving most details to a later date.
Ottawa has said the law known as Bill C-27 would come into force no earlier than 2025.