AI can help in fight against COVID-19, if we don’t let fear stop us: Bengio

TORONTO — One of the godfathers of artificial intelligence says the last year has created a “watershed” moment for the technology, but we have to be careful not to let our fears keep us from exploring it more.

“Innovation is always critical, but I think what’s happened this year is a realization by the public and governments of how much our innovation infrastructure in Canada and throughout the world is important to face the challenges that threaten our economies, our lives like the pandemic,” said Yoshua Bengio.

Bengio is best-known for winning the 2018 Turing Award – nicknamed the Nobel Prize of computing – with Geoffrey Hinton and Yann LeCun, after the trio made a series of deep neural network breakthroughs.

He’s devoted much of his life to researching and advancing AI, which he is hopeful will help in the fight against COVID-19.

Bengio’s remarks come as Mila, the Quebec-based AI research institute he founded in 1993, received a $3.95-million grant from Google Canada.

The grant is spread out over three years and will help support 50 research projects, including LambdaZero.

LambdaZero is using AI to research how to bind something to the SARS-COV-2 peak protein, which the virus uses to infiltrate human cells, and render it inoperative.

The solution could help if vaccines don’t end up working for everyone or if antiviral drugs are needed to ward off death or serious consequences.

But drug discovery is not easy. It costs $2.6 billion per approved drug on average and can take more than a decade in many cases, Mila said. Even when a molecule is found it might not be doing its job very well and might very likely fail in later stages of trials, the institute warns,

“We have been making progress on using machine learning to speed up the search – so that we can do that search more efficiently essentially in a matter of months, rather than years,” Bengio said in an interview.

If successful, LamdbaZero could contribute to saving millions of lives. Testing begins in January.

While the possibilities of AI have been described as endless and touted as the key to future innovation in nearly every industry, the technology has faced a chill.

Many have worried that it will trigger job losses and replace human efforts. Others warn the technology can be weaponized and used to against humanity in a sci-fi-esque scenario.

The concerns have placed heightened attention on privacy and security, which Bengio believes are key to AI’s future.

However, he worries about people having a belief that all AI is troublesome or using those concerns to hold the country back from solving major problems.

He recently had a brush with this attitude when Mila created a contact-tracing app to track probability of infection for users based on their movements and encounters.

Information collected would be anonymously shared with nearby phones to update infection estimations, but the federal government went with a much more pared down app developed by Shopify Inc. employees who volunteered their time, and BlackBerry Ltd.

Mila ended up making its app open-source, so others can use it, but Bengio is still disappointed because his studies showed AI was more accurate in predicting how many people can be infected from one person with COVID-19 than the app the government selected.

“But it’s a political decision and I can understand that governments are prudent,” he said.

“In the days of second wave, I feel even stronger that we collectively didn’t take the right decision, even though I understand it.”

That experience, he said, was insightful and highlighted how much we need to make sure that privacy is maintained through AI but that we don’t let safe opportunities escape us because we don’t fully understand the protections built into it.

“But that’s sort of the unfortunate immediate reaction that happens when technology is used in sensitive places like health care.”