Canada is investing $150M to make AI safer.
asserted
AI → invest → M
Will it work?
asserted
it → work → ?
Lab building AI guardrails gets funding from federal government, Germany
Left unchecked, artificial intelligence could kill us all — that's the stark warning one of the godfathers of the technology, Yoshua Bengio, has been repeating for years.
uncertain
one → build → years
On Wednesday, standing next to representatives of the Canadian and German governments, they together announced over a quarter billion dollars to help the Canadian AI researcher fight that threat.
asserted
researcher → stand → threat
Bengio is trying to create a new kind of AI model.
asserted
Bengio → try → model
The model, called Scientist AI, is being developed to be safe by design: to focus purely on objective truths and understand facts without pursuing goals of its own — like pleasing its human user.
asserted
model → call → user
It could also act as a guardrail for other AI models.
uncertain
It → act → models
The federal government is investing $150 million in LawZero, the Montreal-based non-profit launched by Bengio last year to build models like Scientist AI.
asserted
government → invest → AI
It's the first time that federal funding has gone to the company.
asserted
funding → go → company
Germany is providing an additional 100 million euros.
asserted
Germany → provide → euros
"I'm doing this with the urgency that I feel in my bones, because I want my children and my grandchildren to have a future," said Bengio at the All In conference, a major gathering of AI leaders in Montreal this week.
asserted
Bengio → do → Montreal
Concerns over AI's risks have been mounting, reaching a fever pitch after an outgoing researcher at Anthropic — one of the AI firms considered to be at the forefront of the technology — warned there is a chance AI could wipe out humanity in the next decade.
uncertain
AI → mount → decade
In an interview with Bloomberg earlier this week, Prime Minister Mark Carney also called for a global, non-industry body to oversee and mitigate the technology's risks.
asserted
Carney → call → risks
Read more about LawZero: Can AI safeguard us against AI?
asserted
AI → read → AI
One of its Canadian pioneers thinks so
- Exclusive
Draft federal AI strategy aims to scale up adoption, offer literacy training by 2031
Artificial Intelligence Minister Evan Soloman says the investment in LawZero is part of Canada's larger push for responsible AI development.
asserted
investment → think → development
"We are here today not because we are responding to yesterday's headline, but because we are responding to the philosophic position that we have taken, which is to build responsibly and safely," he said.
asserted
he → respond → that
In many ways, the genie is already out of the bottle, say experts: it will be hugely challenging to curb dangers such as AI hallucinating and spreading misinformation, or taking actions without explicit human permission.
asserted
AI → say → permission
We have very powerful, even transformative, AI systems," said Jaxson Khan, an expert in AI policy and adoption at the University of Toronto's Munk School of Global Affairs.
asserted
Khan → have → Affairs
American and Chinese firms operate incredibly powerful AI systems, a trend he says is likely to continue.
asserted
he → operate → trend
"Canada's at this moment where we're feeling like perhaps, in the context of AI, we don't have a lot of control," he said.
asserted
he → feel → control
But teaming up with Germany to invest in LawZero is still meaningful, Khan says.
asserted
Khan → team → LawZero
"You have to be able to develop alternatives that ideally are safer by design," he said.
asserted
he → have → design
"I'm curious to see if that will work, but I do think it's a worthwhile investment from the government and a good partnership with Germany.
asserted
it → see → Germany
Another important piece of the puzzle, experts say, is partnering with countries that have similar goals and values.
asserted
that → say → goals
The announcement came hours after two AI companies, Canada's Cohere and Germany's Aleph Alpha, signed a merger agreement.
asserted
companies → come → agreement
"This is arguably capacity building," says Vass Bednar, the managing director of the Canadian Shield Institute think-tank, of the merger.
asserted
Bednar → say → 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.
uncertain
One → harm → Canada
"We do have capacity to catch up and build alternatives.
asserted
We → have → alternatives
We also have to believe that," she said.
asserted
she → have → that
"We can't ... fall victim to the inevitability of the technology or its harms.
asserted
We → fall → technology
"
Taking a multipronged approach could make a difference, says Khan.
uncertain
Khan → take → difference
Bengio says giving up isn't an option.
asserted
giving → say → ?
"It won't be easy to achieve our goals," he said.
asserted
he → achieve → goals
"But if we can change the world even by nudging it a little bit towards a better world for humans, we have to try."
asserted
we → change → humans
The Aporia
