June 21, 2024

Advancing Digital Growth

Pioneering Technological Innovation

Nvidia has an AI secret weapon in Canada

13 min read

Handpicked by the AI juggernaut’s CEO to head its Toronto research lab, UofT professor Sanja Fidler and her team are making waves in the crazy-paced world of machine learning

Article content

Sanja Fidler can’t remember exactly what illness kept her home from school for several months as a child, but she remembers the result: upon her return to class after a lengthy absence, she decided, in her little girl heart of hearts, that she did not like math.

This was not a position her mother Soca (pronounced Socha), who was an English teacher in the family’s hometown of Ljubljana, Slovenia, could tolerate, and so began a matriarchal counteroffensive aimed at wooing young Sanja back to the world of numbers by sitting down with the 10-year-old and solving puzzles together.

Advertisement 2

Article content

“Those riddles really got a hold of me,” Fidler said.

Sanja Fidler, as a child, beside her younger sister who she dressed up as a robot for the Slovenian version of Halloween, a costume foreshadowing the career in AI to come
Sanja Fidler, as a child, beside her younger sister who she dressed up as a robot for the Slovenian version of Halloween, a costume foreshadowing the career in AI to come. Photo by Family

Did they ever. The girl who hated math became someone who began bringing her math books on family beach holidays. Other beachgoers lounged on towels and presumably buried their noses in novels, but the self-described math “nerd” solved nerdy math problems, and made a declaration at the grand old age of 12 that she would be studying the subject at university.

That is what Fidler was doing at the University of Ljubljana, circa 2000, when she attended a talk by a visiting professor on computer vision. This was the early days of artificial-intelligence research, when the field was confined to the far reaches of computer science departments at universities and the outer boroughs of geekdom.

Early computer-vision pioneers took visual problems and translated them into math to solve. The young, Slovenian math nerd was mesmerized by the prospects of such technology, so much so that by the time the talk ended, Fidler knew she had found her calling.

“It was love at first sight,” she said. “The fact that you could have something in mathematical form, but then see the results, something that human eyes could understand, and can get excited about, that just felt really nice to me.”

Advertisement 3

Article content

Computer vision has come a long way since Fidler’s transformative moment as an undergrad. It can read X-rays, monitor crops, pilot driverless cars, check out groceries, vacuum floors, recognize faces, keep watch on professional athletes for the analytics gang and generate immersive video-game environments. And that’s just the tip of the AI iceberg, where Fidler today ranks among the premier boundary pushers as the head of Nvidia Corp.‘s research lab in Toronto.

The little girl who hated math grew up to be a brilliant University of Toronto professor and an international star in the computer-vision and machine-learning world. Her team’s work in Toronto has given creative types the superpowers to instantly generate entire 2D and 3D scenes — a mountain lake shrouded in fog, a woman beneath a vast Prairie sky or any other visual that comes to mind — simply by prompting the AI with text. Other research supports Nvidia’s autonomous vehicle technologies that are wrapped into the company’s ongoing collaboration with Mercedes-Benz AG.

This is incredibly serious stuff, but Fidler is not without a sense of humour. As a professor, she occasionally bridged the divide separating the experts from mere ordinary folks to showcase her AI chops. This included developing a karaoke-singing neural network that could look at a photograph and generate a song. The big hit to emerge from it was a Christmas carol inspired by an image of a Christmas tree that briefly became an internet sensation and wound up on an episode of The Simpsons.

Advertisement 4

Article content

“I think they were making fun of AI,” Fidler said. “I should put it on my CV.”

No need. Jensen Huang, co-founder and chief executive of Nvidia — the juggernaut whose technology is powering AI advances that can be found, for example, at the machine-learning heart of generative AI wonders such as OpenAI OpCo LLC’s ChatGPT — handpicked Fidler in 2018 to head up its Toronto lab. Just imagine being handpicked by Santa Claus to guide his sleigh at night, and you get a sense of the magnitude of the assignment.

Article content

Nvidia co-founder and chief executive Jensen Huang handpicked Sanja Fidler in 2018 to head up the company’s Toronto lab.
Nvidia co-founder and chief executive Jensen Huang handpicked Sanja Fidler in 2018 to head up the company’s Toronto lab. Photo by David Paul Morris/Bloomberg

“I like solving problems,” Fidler said. “In academia, I feel like we are inventing problems, but industry has problems, so that real-world aspect of it really drew me in.”

Rise of the Magnificent Seven

Nvidia wasn’t just another tech startup when Fidler came aboard as employee No. 1 at the California-based company’s Canadian research outpost. The company had been knocking around since the 1990s, and it was well-known among video gamers for its graphics processing units (GPUs) that made video-game graphics pop with high-resolution imagery.

What wasn’t really known until Geoffrey Hinton, a.k.a. the Godfather of AI, Alex Krizhevsky and Ilya Sutskever scored a major breakthrough in image recognition at the U of T in 2012 was that GPUs were key to unlocking artificial intelligence’s magic.

Advertisement 5

Article content

Huang discerned the magic, and Nvidia plowed a bunch of money into research and development. Lo and behold, when ChatGPT launched in November 2022, companies, governments and working folks everywhere suddenly saw with their human eyes that generative AI was not just an abstract moonshot the eggheads were working on, but something that could transform life and work as we know it. By then, Nvidia had already established itself as a dominant global player for the AI-related hardware and software that was driving the revolution.

Geoffrey Hinton, known as the Godfather of AI, helped score a major breakthrough in image recognition at the University of Toronto in 2012
Geoffrey Hinton, known as the Godfather of AI, helped score a major breakthrough in image recognition at the University of Toronto in 2012 Photo by GEOFF ROBINS/AFP via Getty Images

Fast-forward to today and the company has 30,000 employees and shares that trade for US$800 a pop, give or take — or more than double the share price of a year ago, and 1,745 per cent greater than five years ago. Nvidia’s US$2-plus trillion market capitalization has put the company in rarefied air alongside Apple Inc. and Microsoft Corp. as the three most valuable publicly traded corporations on the planet, while Amazon.com Inc., Google owner Alphabet Inc., Meta Platforms Inc. and Tesla Inc. have also plowed billions into AI technologies. Together, they make up the so-called Magnificent Seven that everyone from geeks to investors are obsessed with.

“Nvidia is defining the market, and everybody else is playing catch-up, and they are playing catch-up against a moving target,” Stacy Rasgon, a senior technology analyst at Bernstein Research, said.

Advertisement 6

Article content

He has been on the job for 15 years and said he has never seen a company’s fortunes go on a bender quite like Nvidia’s have over the past few years. A big part of what is pushing the company’s success is Huang’s continued commitment to research and development, and being the first to come out with what comes next, and that’s where Fidler and her crew come in.

On a Tuesday in early April, the vice-president of AI research at Nvidia was on the U of T campus gearing up for final exams. Fidler got the job at Nvidia, but kept her university job, and she teaches an introductory course on computer vision. Keeping one foot in the classroom was a must-have in her negotiations with Huang. Asked whether being compensated in Nvidia shares was similarly a must-have and she said, “Don’t worry, I am good.”

Fidler initially met Huang at a computer-vision conference in Hawaii right after she had won an award for a paper. As pennywise academics are inclined to do, she elbowed up to the open bar at a reception Nvidia was hosting when she fell into a conversation with her future boss.

This was the era when tech giants were actively poaching AI experts from universities, including Hinton, who had gone to work for Google. Nvidia wasn’t the only private-sector outfit Fidler was speaking to, but Huang was the only CEO who personally followed up with a phone call to tell her the job in Toronto was hers, and that she could keep her students, a handful of whom now work for the company.

Advertisement 7

Article content

Article content

Nvidia CEO Jensen Huang speaks with Elon Musk during the GPU Technology Conference (GTC) in San Jose, California in 2015, when Musk said that we’ll “take autonomous cars for granted” in a short period of time.
Nvidia CEO Jensen Huang, left, speaks with Elon Musk during the GPU Technology Conference (GTC) in San Jose, California in 2015. Musk said at the time that we’ll soon “take autonomous cars for granted.” Photo by David Paul Morris/Bloomberg

“I think Jensen has a lot of respect for Toronto because of this AI boom; he has a special place in his heart for it,” Fidler said. “It felt like the right opportunity to me, and he was very receptive to working with students and creating this AI presence here. And he was a cool guy, which really resonated with me.”

Don’t be fooled by all the math-nerd stuff. Fidler possesses a cool factor of her own. She is tall, lean and athletic-looking. She played basketball for a year in her youth, but sports didn’t take. Instead, she thrived in crushing all-comers at math competitions.

Employee No. 1 at the research lab now has a team of about 35 researchers working for her, and she expects to hit a headcount of 40 soon. Nvidia can clearly afford to spend on the expertise, given the company’s outlay on research and development in fiscal 2024 was US$8.68 billion.

AI is more than tech

But money isn’t everything. Arguably, the lab’s greatest selling point has been the person who runs it. Undeniably smart, Fidler is also extremely likable. She asks questions, takes an interest in others and is open, something that seems to have given Nvidia a recruiting advantage in what has become a global arms race for AI talent.

Advertisement 8

Article content

“I know that for myself, and many of my colleagues, Sanja’s leadership was one of the main factors in joining the lab,” Masha Shugrina, a senior research scientist and manager at Nvidia in Toronto, said. “In addition to her brilliance and creativity, she values and nurtures everyone’s ideas.”

In other words, Fidler gets it. Research can be a brutal grind, especially for young people, since “most of the time, nothing works,” she said. Researchers are chasing the unknown, and looking to solve problems that no one has ever solved before. It is a risky business, fraught with failure and the nagging voice of self-doubt.

“There is a lot of imposter syndrome,” Fidler said. “In that kind of environment, you want to be encouraging, you want to be kind, and you want to make sure that people feel as though you are in this together.”

The all-for-one and one-for-all leadership style was evident long before Fidler made the leap to the private sector. Sven Dickinson is another U of T professor, and there happens to be a lot of them, who were poached by the private sector — in his case, Samsung Electronics Co. Ltd. grabbed him to head up an AI research hub in Toronto.

He first caught wind of Fidler’s work in 2006, when she published a paper on object recognition that predated the deep-learning revolution to come. He later invited her to join his research group at the university prior to Fidler being offered a full-time gig as a professor.

Advertisement 9

Article content

“Sanja’s leadership in the academic community is remarkable,” Dickinson said.

Equally remarkable, he said, has been her leadership in the industrial sector, but perhaps most remarkable of all, is that despite being an international AI superstar — granted, one most ordinary people, including a group of U of T students quietly working away in pairs in a sunlit space on campus with the luminary in their midst would not recognize — Fidler hasn’t changed.

She is still Sanja, and while she may possess the means to afford a fancy car, the make of which she was too embarrassed to disclose, as well as own a home in a pricey neighbourhood near campus — and have a massive, stressful job at a massive, global tech company — she continues to be among the “kindest and most sensitive” colleagues Dickinson has ever run across.

But being kind will only get you so far in the private sector. Unlike university professors, senior executives don’t have the security of tenure, and they can be fired at any time.

Ultimately, they need to produce.

“Before, I could kill a project,” Fidler said. “There would be a bad mood, but now, you can’t really kill a big project with 20 and 30 people on it, so you are meant to deliver, especially because there are a lot of resources.”

Advertisement 10

Article content

Research can be a brutal grind, especially for young people, since “most of the time, nothing works,” says Fidler.
Research can be a brutal grind, especially for young people, since “most of the time, nothing works,” says Fidler. Photo by Peter J. Thompson/National Post

Generation AI

In the early days of AI, only a small handful of supersmart people were mulling what were then big ideas, such as how a computer might learn to correctly identify — or, if you will see — a cat as a cat, instead of a dog, in an image. That may not sound like anything of great note, but it becomes radically important when autonomous vehicles need to be able to spot the difference between an empty stretch of road and a road with the pedestrian in the middle of it.

The emergence of ChatGPT has introduced generative AI to the mainstream rabble. Parents stumped for birthday party ideas for that six-year-old kid in their lives can now ask AI to generate some birthday party recommendations for them and, presto, it will in a pinch.

But generative AI has been Fidler and Co.’s “bread and butter” at the lab since day one. And what has her jazzed and a wee bit stressed these days is whether she and her team can push the machine-learning boundaries to a place where, say, you type a word and AI renders entire immersive virtual environments.

Imagine a beach scene complete with dolphins, sand and sun. And imagine, as the human in virtual-reality goggles moving through the scene, being able to smell the salt-air breeze, actively interact with objects in the environment and, who knows, maybe even hitch a lift on a dolphin.

Advertisement 11

Article content

“The frontier is pushing the scale of the environment, the realism of it,” she said. “This technology is typically perceived for creative purposes — I want to generate a new movie, or whatever — but you can also take this technology and use it to drive a car or drive a humanoid robot — again, that is our belief.”

Delivering upon that belief could add another lucrative dimension to a company that reported US$60.9 billion in revenue in fiscal 2024, but just one-60th of which was derived from its automotive sector.

A 2023 report from McKinsey & Co. predicts the market for AI-driven “advanced driver assistance systems and autonomous-driving systems” could be worth as much as US$400 billion by 2035, suggesting there is a massive opportunity at stake for an AI giant, as well as a collaboration with Mercedes-Benz to make good on.

But don’t fret too much for Nvidia; the little girl who did not like math is on the case.

“I am a problem-solver,” Fidler said. “Any problem you throw at me, I am going to be thinking about it.”

• Email: [email protected]

Bookmark our website and support our journalism: Don’t miss the business news you need to know — add financialpost.com to your bookmarks and sign up for our newsletters here.

Article content

link

Leave a Reply

Your email address will not be published. Required fields are marked *

Copyright © All rights reserved. | Newsphere by AF themes.