This November, Amii Fellow and Canada CIFAR AI Chair Adam White organized the University of Alberta’s inaugural AI Day. The event featured AI experts who discussed the impact of artificial intelligence on scientific research and highlighted some of the groundbreaking research being conducted in Edmonton.

AI in Science

The first discussion of the event featured two Amii Fellows and Canada CIFAR AI Chairs, Martha White and Abby Azari, as well as biochemist Glen Uhrig, principal investigator of the U of A’s Uhrig Lab. The trio discussed how machine learning is having a transformational effect on scientific research and is leading to promising solutions to weighty, real-world problems.

Uhrig, whose lab studies how the biology of plants changes throughout the day and how they are affected by different lighting and environmental conditions, started by explaining that he was the panelist with the least experience in machine learning.

“I used to think it was the stuff of Hollywood movies, “ Uhrig said to open the panel. But now that he is more familiar with machine learning, he sees it differently.

“It’s a way to get to new ideas. It’s a way to accelerate discovery.”

Urhig says his lab has started to use machine learning to analyze images of plants that are grown under different lighting conditions. The models they are using are able to pick up subtle changes in the plants that might be missed by the human eye, allowing them to test the effects of different growing conditions.

Urhig says that this is only in the early stages. He thinks artificial intelligence has the potential to support exciting advancements in plant science and biochemistry, leading to techniques to help crops grow faster and produce more. This could have huge implications for indoor farming, which could be extremely important for growing food in remote northern communities or even for food production during space travel.

Abby Azari offered more examples of where machine learning might support space exploration. An Assistant Professor jointly appointed in the departments of Physics and Electrical and Computer Engineering at the U of A,  Azari’s work focuses on using AI to reveal the secrets of the solar system.

She notes that machine learning has become indispensable for space exploration. The challenges of analyzing the data collected during space exploration are somewhat contradictory: sometimes, it is an issue of too much data, and sometimes it is not enough. She says machine learning proves valuable in dealing with both problems; it allows researchers to make use of the nearly endless streams of complex data about other planetary bodies collected from rovers and satellites, while also helping fill in the gaps in data for rare astronomical events, helping space scientists build a better understanding of them.  

When asked what types of future applications she was most excited about, the researcher said machine learning would be vital when it came to tasks like examining the potential of oceans hidden under the surface of other planets and moons in the solar system.

“You really need ML to do this. There is a lot of uncertainty there.”

Martha White provided a more grounded example of how AI is having an impact on the water here on Earth. Much of White’s research is dedicated to building robust, durable reinforcement learning systems that can handle the complexity of the real world. White talked about the AI projects being led by RLCore Technologies, the company she co-founded.

The company is using reinforcement learning to help control and optimize operations in industrial settings, primarily in water treatment, leading to better efficiency and a reduced environmental impact. She says that while automation has been a part of industrial operations for years, AI’s ability to respond to feedback and adjust for changing conditions makes it invaluable in the complex, dynamic environments industrial control systems work in.

To close out the panel, White also addresses a question on whether there was a danger of artificial intelligence systems replacing human scientists completely.

“I think that idea is very rare,” she said. “Science is done as a team. AI in science is usually about empowering people to do things differently and allowing them to do new things they couldn’t do before.”

Careers in AI

AI Day’s second panel discussion switched the conversation from research to career advice. The panelists of three Amii Fellows and Canada CIFAR AI Chairs — Marlos C. Machado, Bei Jiang, and Alona Fyshe — offered advice and insight for computer science students as to their options after graduation.

Much of the conversation centred around the question facing any graduating student: whether to look for a career in industry or academia. The three panelists offered a mix of different experiences in that regard. Both Machado and Fyshe have careers that were a mix of industry and academic jobs, working for companies including Microsoft, Google, and IBM. Jiang, on the other hand, has remained in academia.

The panelists noted that the barriers between academic and industry careers have weakened significantly in recent years, and switching between the two is no longer as difficult as it once was. Although Machado did note that decisions to move between the two paths should be made intentionally.

“You can switch between the two, yes,” he said. “But you always have to make sure whatever you do is to make yourself marketable.”

Similar advice was given when it came to the question of gaps in one’s work experience. They aren’t necessarily devastating when it comes to being hired, but they argued that any gaps should have a reason for it. Fyshe noted that employers in 2025 are much more understanding that personal or family reasons might take people out of the workforce temporarily, but those gaps often need to be explained. Jiang responded that the attitudes towards gaps can also be very different between different industries and disciplines. She pointed out that in her field, statistics, gaps in employment aren’t uncommon, and can sometimes even be a positive if it was due to interesting research or an impactful project.

In the end, the trio of researchers all agreed on one thing: there is no single path to happiness. There are endless paths that a successful career can take, and the most important thing is to follow the areas that you find most interesting. Following that instinct is what led all three of the panelists to jobs that they find fulfilling.

“I had a drive to understand,” Jiang said about what led her to her current position. “I followed that curiosity.”

Rich Sutton: The Era of Experience

In the final keynote of the day, Amii’s Chief Scientific Advisor and recent Turing Award winner, Rich Sutton, outlined what he saw as a profound shift in the future of artificial intelligence in the near future.

Sutton argues that until now, we have been living in what he calls the Era of Human Data. This is exemplified by the rise of Large Language Models (LLMs), which are trained on massive datasets made up of data created by humans: videos, books, articles, databases, and other media that have been created and collected by human beings. And while this has led to impressive achievements by LLMs, Sutton says we are close to reaching the limit for groundbreaking leaps forward in this area. But by remaining limited by human data and imitating human modes of reasoning, they cannot progress past the scope of human knowledge.

“The entire purpose of AI right now is to transfer human knowledge to these Large Language models,” he said. “And I find that really strange.”

To truly achieve the potential that AI holds, Sutton says we must move past the Era of Human Data and into the Era of Experience: one where artificial intelligence agents learn through continual interaction with their environment through reinforcement learning. But getting to that level will require a greater understanding of improving methods that allow models to learn continually. Learning from experience cannot be done in the large, discrete training sessions that power current LLM development. Instead, AI will need to have the ability to incorporate a continuous stream of knowledge to change the way it interacts with the world, the same way that organic intelligences do.

Sutton points out that leaving the era of human data will likely lead to artificial intelligence that has different methods of reasoning than human beings and might also have different goals. Instead of trying to eliminate that possibility, Sutton argues for an approach that emphasizes cooperation, which he calls “humanity’s superpower,” and to avoid the temptation to attempt centralized control of artificial intelligence. A movement towards the Era of Experience is a path towards a fuller understanding of what intelligence really is, which Sutton argues would be the most important scientific discovery in human history.

​“Intelligence is the most powerful phenomenon in the universe,” Sutton said.

Want to learn more about Rich Sutton's thoughts on the Age of Experience? Check out his presentation from Upper Bound 2025.