Artificial Intelligence

Geoffrey Hinton

Nobel Prize

Physics

Subheadline

A U of T University Professor Emeritus, Hinton shared the honour with Princeton University's John J. Hopfield for discoveries and inventions that enable machine learning with artificial neural networks

Geoffrey Hinton, a University Professor Emeritus of computer science at the �ؿ�ζSM, has won the 2024 Nobel Prize in Physics.

Widely regarded as the “godfather of AI,” Hinton shared the prize with John J. Hopfield of Princeton University for foundational discoveries and inventions that enable machine learning with artificial neural networks.

Hinton said he was “flabbergasted” at the honour as messages poured in from around the world.

“I had no expectations of this,” he told U of T News shortly after the win was announced in Stockholm Tuesday morning. “I am extremely surprised and I'm honoured to be included.”

He later told reporters at a press conference he was “in a cheap hotel in California” with no Internet and a poor phone connection when he was notified about his Nobel Prize.

“I was going to get an MRI scan today, but I think I’m going to have to cancel that.”

Hinton and Hopfield are credited with wielding tools from physics to advance basic research in the field. Specifically, Hopfield created an associative memory that can store and reconstruct images in data, while Hinton invented a way to find properties in data and perform tasks such as identifying specific elements in pictures.

“On behalf of the �ؿ�ζSM, I am absolutely delighted to congratulate University Professor Emeritus Geoffrey Hinton on receiving the 2024 Nobel Prize in Physics,” said U of T President Meric Gertler. “The U of T community is immensely proud of his historic accomplishment.”

Hinton was selected for the high-profile award for his use of the Hopfield network – invented by his co-laureate – as the foundation for a new network called the Boltzmann machine that can learn to recognize elements within a given type of data.

The Boltzmann machine can classify images and generate new examples of the pattern on which it was trained, with Hinton and his graduate students later building on this work to help usher in today’s rapid development of machine learning – a technology that now underpins a host of applications ranging from large language models such as ChatGPT to self-driving cars.

Johan Jarnestad/The Royal Swedish Academy of Sciences

“The laureates’ work has already been of the greatest benefit. In physics we use artificial neural networks in a vast range of areas, such as developing new materials with specific properties,” said Ellen Moons, chair of the Nobel Committee for Physics.

The win by Hinton and Hopfield was covered by media and other organizations around the globe, with The New York Times describing the Nobel committee’s decision as “an acknowledgement of AI’s growing significance in the way people live and work,” and the prestigious journal Nature noting Hinton’s innovations now “form the basis of many state-of-the-art AI tools.”

Hinton joined U of T as a professor of computer science in 1987 after working in various universities in the U.K., where he was born, and in the United States. He went on to be named a University Professor – U of T’s highest academic appointment – in 2006.

Driven by a desire to understand the human brain, Hinton and his graduate students built on his early efforts with an array of developments that paved the way for an explosion in deep learning. One of the first cohort of researchers supported by the Canadian Institute for Advanced Research (CIFAR), Hinton’s work helped catapult Canada to its current status as a global leader in AI development.

The Royal Swedish Academy of Sciences, which awards the Nobel Prize in Physics, noted Hinton persisted with his research even as the scientific community lost interest in artificial neural networks during the 1990s, and ultimately “helped start the new explosion of exciting results” in the 2000s.

Hinton, for his part, said during a U of T press conference Tuesday evening that his achievements wouldn’t have been possible without support for curiosity-based research – something he said Canada was good at.

He added that his shock at winning the Nobel stemmed from the fact that, while his work has drawn on statistical physics, he isn’t a physicist himself – and even “dropped out of physics after my first year in university because I couldn’t do the complicated math.”

He also said that he plans to donate the money associated with the prize to various charities, including one that provides jobs for neurodiverse young adults.

Hinton likened the influence of AI to that of the Industrial Revolution during a virtual press conference with the academy earlier in the day – “But instead of exceeding people in physical strength, it’s going to exceed people in intellectual ability.”

He added that the rise of AI “is going to be wonderful in many respects,” citing health care and workplace productivity as two areas poised to benefit hugely from the technology. “But we also have to worry about a number of possible bad consequences, particularly the threat of these things getting out of control,” Hinton said.

In early 2023, Hinton quit his job at Google and focused on sounding the alarm about the risks of rapid and unfettered AI development. He outlined his reasoning in a 46-minute U of T video last year, urging young researchers to focus their efforts on the emerging field of AI safety – a message he repeated in media interviews following his Nobel win.

He has continued to tackle the issue at lectures and public appearances around the world, including at U of T and at Cambridge University, his alma mater.

“I am thrilled Geoffrey Hinton, an esteemed colleague and dear friend has been awarded the Nobel Prize in Physics,” said Melanie Woodin, dean of U of T’s Faculty of Arts & Science.

“Geoff is an historic visionary whose groundbreaking work in deep learning and neural networks has made U of T and the Toronto region a leading global centre for AI. And it speaks volumes about his integrity that while he helped lay the foundation for the artificial intelligence revolution, he is also one of the leading voices urging that we develop this technology responsibly and ethically.”

Similarly, Prime Minister Justin Trudeau lauded Hinton for his efforts to realize responsible AI development, releasing a statement and writing on X: “Geoffrey, we’re glad to have a mind like yours developing safe and responsible AI for the world.”

Geoffrey Hinton delivers a lecture about responsible AI to U of T students and faculty (photo by Johnny Guatto)

Hinton, who is co-founder and chief scientific adviser at the Vector Institute in Toronto, joins an illustrious list of past Nobel Prize in Physics winners that includes Albert Einstein and Marie Curie (who also won a Nobel in chemistry). The prestigious award is the latest in a long list of accolades for Hinton. They include the Association for Computing Machinery’s A.M. Turing Award – widely considered “the Nobel Prize of computing” – in 2019 alongside collaborators Yann LeCun and Yoshua Bengio.

Hinton is the fourth U of T faculty member to win a Nobel Prize over the years.

Sir Frederick Banting and J.J.R Macleod won a Nobel Prize in Physiology or Medicine for their work with Charles Best in 1923 to isolate insulin. In 1986, John Polanyi was one of three winners of the Nobel Prize in Chemistry for the development of the new field of reaction dynamics.

Other members of the U of T community, including several alumni, have received or been associated with the international honour.

Oliver Smithies, a past professor at U of T, was a joint winner of the Nobel Prize in Physiology or Medicine in 2007 for discovering the “principles for introducing specific gene modifications in mice by the use of embryonic stem cells.”

Also in 2007, Professor Robert Jefferies shared in the Nobel Peace Prize awarded to the Intergovernmental Panel on Climate Change (IPCC), in which he was a key Canadian representative as an international leader in arctic science and global change biology.

In 1999, U of T Professor James Orbinski accepted the Nobel Peace Prize on behalf of Doctors Without Borders, which was recognized for its humanitarian work.

Anti-nuclear activist and U of T alumna Setsuko Thurlow accepted the Nobel Peace Prize in Norway in 2017 on behalf the International Campaign to Abolish Nuclear Weapons (ICAN).

In 2001, Michael Spence, an alumnus of �ؿ�ζSM Schools, was one of three joint winners of the Bank of Sweden Prize in Economic Sciences in Memory of Alfred Nobel for his contributions to analyses of markets with asymmetrical information.

Bertram Brockhouse, who completed two degrees at U of T, was a co-winner of the Nobel Prize in Physics in 1994 for the development of neutron scattering techniques for studies of condensed matter.

Arthur Schawlow, an alumnus, was one of three winners of the same prize in 1981 for his contribution to the development of laser spectroscopy.

In 1998, U of T alumnus Walter Kohn was a co-winner of the Nobel Prize in Chemistry for development of the density-functional theory.

Former Prime Minister Lester B. Pearson, who received a bachelor’s degree from U of T, won the Nobel Peace Prize in 1957.

U of T experts tackle questions about AI safety, ethics during panel discussion

rahul.kalvapalle — Wed, 02 Oct 2024 18:33:43 +0000

U of T experts tackle questions about AI safety, ethics during panel discussion

rahul.kalvapalle Wed, 10/02/2024 - 14:33

Cutline

From left: U of T's Roger Grosse, Sedef Kocak, Sheila McIlraith and Karina Vold take part in a panel discussion on AI safety (photo by Duane Cole)

Kyle Coulter

Jovana Jankovic

Topic

Schwartz Reisman Institute for Technology and Society

The Institute for the History and Philosophy of Science and Technology

Department of Computer Science

Centre for Ethics

Department of Philosophy

Vector Institute

Victoria College

Subheadline

"We should be building AI systems that promote human flourishing – that allow human beings to live with dignity and purpose, and to be valued contributors to society”

What does safe artificial intelligence look like? Could AI go rogue and pose an existential threat to humanity? How have media portrayals of AI influenced people’s perceptions of the technology’s benefits and risks?

These were among the pressing questions tackled by four experts at the �ؿ�ζSM and its partner institutions – in disciplines ranging from computer science to philosophy – during a recent panel discussion on AI safety.

Sheila McIlraith, professor in U of T’s department of computer science at the Faculty of Arts & Science and Canada CIFAR AI Chair at the Vector Institute, said the notion of AI safety evokes different things to different people.

“Computer scientists often think about safety critical systems – the types of systems that we’ve built to send astronauts to the moon or control our nuclear power plants – but AI safety is actually quite different,” said McIlraith, an associate director at the U of T’s Schwartz Reisman Institute for Technology and Society (SRI).

“For me personally, I have a higher bar, and I really think we should be building AI systems that promote human flourishing – that allow human beings to live with dignity and purpose, and to be valued contributors to society.”

The event, hosted by SRI in partnership with the Vector Institute, the Institute for the History & Philosophy of Science & Technology, the Centre for Ethics and Victoria College, invited McIlraith and her fellow panelists to discuss how AI technologies can be aligned with human values in an increasingly automated world.

They also discussed how risks surrounding the technology can be mitigated in different sectors.

Karina Vold, the event’s moderator, underscored the challenge of building safe AI systems in an uncertain world (photo by Duane Cole)

Moderator, Karina Vold, assistant professor in the Institute for the History & Philosophy of Science & Technology in the Faculty of Arts & Science, noted that because AI systems operate “in a world filled with uncertainty and volatility, the challenge of building safe and reliable AI is not easy and mitigation strategies vary widely.”

She proceeded to ask the panel to share their thoughts on the portrayal of AI in popular culture.

“The media devotes more attention to different aspects of AI – the social, philosophical, maybe even psychological,” said Sedef Kocak, director of AI professional development at the Vector Institute.

“These narratives are important to help show the potential fears, as well as the positive potential of the technology.”

The discussion touched on several topics related to AI safety (photo by Duane Cole)

Roger Grosse, associate professor in U of T’s department of computer science in the Faculty of Arts & Science and a founding member of the Vector Institute, said that safety concerns around AI are not merely rooted in science and pop culture, but also in philosophy.

“Many people think that the public’s concerns regarding AI risks come from sci-fi, but I think the early reasoning regarding AI risks actually has its roots in philosophy,” said Grosse, who also holds Schwartz Reisman Chair in Technology and Society.

“If we’re trying to reason about AI systems that don’t yet exist, we don’t have the empirical information, and don’t yet know what their design would be, what we can do is come up with various thought experiments. For example, what if we designed an AI that has some specific role, and all of the actions that it takes are in service of the role?

“For the last decade, a lot of the reasons for being concerned about the long-term existential risks really came from this careful philosophical reasoning.”

The discussion also touched on the dangers of AI models misaligning themselves, how to guard against bias in the training of large language models, and how to ensure that AI models with potentially catastrophic capabilities are safeguarded.

“This [safeguarding] is an area where new research ideas and principles will be required to make the case,” said Grosse. “Developers saying, ‘Trust us’ is not sufficient. It’s not a good foundation for policy.”

Despite addressing topics surrounding potential harms and risks of AI, the panelists also shared their optimism about how AI can be wielded for the greater good – with Grosse noting AI offers the promise of making knowledge more widely accessible, and Kocak focusing on the myriad benefits for industries.

Watch the Sept. 10 conversation below:

AI tool reduces risk of unexpected hospital deaths by 26 per cent: Study

Christopher.Sorensen — Mon, 23 Sep 2024 16:50:13 +0000

AI tool reduces risk of unexpected hospital deaths by 26 per cent: Study

Christopher.Sorensen Mon, 09/23/2024 - 12:50

Cutline

(photo by Tempura/Getty Images)

Adam Miller

Topic

Temerty Faculty of Medicine

Unity Health

Temerty Faculty of Medicine

St. Michael's Hospital

Subheadline

CHARTWatch, an artificial intelligence early warning system developed at Unity Health Toronto, alerts physicians when hospital patients face a high risk of unexpected death or transfer to an ICU

A new study by Toronto researchers evaluates the use of CHARTWatch, an artificial intelligence early warning system developed at Unity Health Toronto that monitors hospitalized patients in real-time, identifies those at high risk of unexpected death or transfer to an intensive care unit, and alerts doctors and nurses to intervene early.

The study, published in the Canadian Medical Association Journal, shows a 26 per cent reduction in unanticipated mortality after the tool was implemented in the general internal medicine ward of Unity Health Toronto’s St. Michael’s Hospital.

“As AI tools are increasingly being used in medicine, it is important that they are evaluated carefully to ensure that they are safe and effective,” says lead author Amol Verma, general internist at Unity Health and professor in the department of medicine in the �ؿ�ζSM’s Temerty Faculty of Medicine who led the development and implementation of CHARTWatch.

Amol Verma (L) and Muhammad Mamdani (supplied images)

“Our findings suggest that AI-based early warning systems are promising for reducing unexpected deaths in hospitals.”

One of the primary sources of unplanned admission to the ICU is the unexpected deterioration of hospitalized patients, which prompted the research team to develop this AI tool and study its effectiveness.

“This important study evaluates the outcomes associated with the complex deployment of the entire AI solution, which is critical to understanding the real-world impacts of this promising technology,” says study co-author Muhammad Mamdani, vice president of data science and advanced analytics at Unity Health and director of U of T’s Temerty Centre for Artificial Intelligence Research and Education in Medicine.

“We hope other institutions can learn from and improve upon Unity Health Toronto’s experiences to benefit the patients they serve.”

The study analyzed data from 13,649 patients aged 55 to 80 years old admitted to the general internal medicine unit (9,626 in the pre-intervention period and 4,023 using CHARTWatch) and 8,470 admitted to subspeciality units that did not use CHARTWatch.

Helping to prioritize patient needs

“The CHARTWatch project started at Unity Health when we asked patients, clinicians, hospital leaders, what would you want to use artificial intelligence for? If you could predict one thing that AI would tell you, what should that be?” says Verma, who is also the Temerty Professor of AI Research and Education in Medicine.

“And one of the leading priorities of everyone was to be able to predict in advance which patients might become so sick in hospital that they need ICU or might die.”

During the 19-month-long intervention period, 482 patients in the general internal medicine became high-risk, compared with 1,656 patients who became high risk in the 43-month-long pre-intervention period. There were also fewer non-palliative deaths in the CHARTWatch group than in the pre-intervention group (1.6 per cent versus 2.1 per cent).

“If I were a patient, I would be so relieved to know that we have this kind of system,” said co-author Yuna Lee, division head and general internist at St. Michael’s and professor in the department of medicine at U of T.

“So, when the patient gets a high-risk alert, they’re going to be assessed by senior staff right away. Also, they’re going to have quite close monitoring by nursing staff, so they check on them every one to two hours.”

Expanding the impact of CHARTWatch

CHARTWatch inputs more than 100 aspects of a patient’s medical history and current health status that are routinely stored in the hospital’s electronic medical record. It analyzes the interactions between these inputs and how they change over time. With that information, it’s able to categorize each patient by their risk for deterioration and send an alert to prioritize treatment.

The study is one of the first to evaluate how CHARTWatch can benefit hospital patients in Canada and shows the potential real-world impact of AI on the health-care sector. Damian Jankowicz, Unity Health’s vice-president and chief information and AI officer, says he hopes AI tools such as CHARTWatch will continue to have a profound impact on patients.

“Hopefully with reduced administrative burden on our providers, they will have more time to spend with our patients and really focus on the patient needs,” he says.

“I hope that AI will be able to distill the incredible amounts of information that’s coming at our clinicians into important components and really bring their clinical judgment to the forefront.”

This story first was first published by Unity Health.

U of T researchers develop AI model to predict 'very dynamic' peptide structures

Christopher.Sorensen — Thu, 15 Aug 2024 12:54:41 +0000

U of T researchers develop AI model to predict 'very dynamic' peptide structures

Christopher.Sorensen Thu, 08/15/2024 - 08:54

Cutline

PhD Graduate Osama Abdin and Professor Philip M. Kim developed a deep-learning model that can predict all possible shapes of peptides, which are are of keen interest to researchers who are developing therapeutics (supplied image)

Anika Hazra

Topic

Breaking Research

Donnelly Centre for Cellular & Biomolecular Research

Subheadline

The new model expands on the capabilities of Google DeepMind's AlphaFold, the leading AI system for predicting protein structures

Researchers at the �ؿ�ζSM have developed a deep-learning model that can predict all possible shapes of peptides – chains of amino acids that are shorter than proteins, but perform similar biological functions.

Called PepFlow, the model combines machine learning and physics to model the range of folding patterns that a peptide can assume based on its energy landscape.

Peptides, unlike proteins, are dynamic molecules that can take on a range of conformations. They are involved in many biological processes that are of keen interest to researchers who are developing therapeutics.

“We haven’t been able to model the full range of conformations for peptides until now,” said Osama Abdin, first author on the study and recent PhD graduate of molecular genetics at U of T’s Donnelly Centre for Cellular and Biomolecular Research. “PepFlow leverages deep-learning to capture the precise and accurate conformations of a peptide within minutes.

“There’s potential with this model to inform drug development through the design of peptides that act as binders.”

The study was recently published in the journal Nature Machine Intelligence.

A peptide’s role in the human body is directly linked to how it folds since its 3D structure determines the way it binds and interacts with other molecules.

“Peptides were the focus of the PepFlow model because they are very important biological molecules and they are naturally very dynamic, so we need to model their different conformations to understand their function,” said Philip M. Kim, the study’s principal investigator and a professor at the Donnelly Centre. “They’re also important as therapeutics, as can be seen by the GLP1 analogues, like Ozempic, used to treat diabetes and obesity.”

Peptides are also cheaper to produce than their larger protein counterparts, said Kim, who is also a professor of computer science in U of T’s Faculty of Arts & Science and a professor of molecular genetics in the Temerty Faculty of Medicine.

The new model expands on the capabilities of AlphaFold, the leading Google DeepMind AI system for predicting protein structure. It does this by generating a range of conformations for a given peptide. Taking inspiration from highly advanced physics-based machine learning models, PepFlow can also model peptide structures that take on unusual formations, including the ring-like structure that results from a process called macrocyclization. Peptide macrocycles are currently a highly promising venue for drug development.

“It took two-and-a-half years to develop PepFlow and one month to train it, but it was worthwhile to move to the next frontier beyond models that only predict one structure of a peptide,” Abdin said.

There are, however, limitations given that PepFlow represents the first version of a new model. The study authors noted a number of ways in which PepFlow could be improved, including training the model with explicit data for solvent atoms, which would dissolve the peptides to form a solution, and for constraints on the distance between atoms in ring-like structures.

Yet, even as a first version, the researchers say PepFlow is a comprehensive and efficient model with potential for furthering the development of treatments that depend on peptide binding to activate or inhibit biological processes.

“Modelling with PepFlow offers insight into the real energy landscape of peptides,” said Abdin.

The research was supported by the Canadian Institutes of Health Research and the Natural Sciences and Engineering Research Council of Canada.

U of T team takes top spot in self-driving car challenge for 6th time in 7 years

Christopher.Sorensen — Wed, 07 Aug 2024 17:42:55 +0000

U of T team takes top spot in self-driving car challenge for 6th time in 7 years

Christopher.Sorensen Wed, 08/07/2024 - 13:42

Cutline

As part of the competition, the U of T team’s autonomous vehicle had to react to obstacles such as a fake deer moving across the road (photo courtesy of aUToronto)

Safa Jinje

Topic

Faculty of Applied Science & Engineering

Graduate Students

Self-Driving Cars

Undergraduate Students

UTIAS

Subheadline

"Each time we saw an obstacle – a stop sign, a red light, the railroad bar coming down – and the car reacted by stopping and then continuing, we let out a big cheer or a sigh of relief"

A team from the �ؿ�ζSM has placed first for sixth time in seven years in a North American self-driving car competition.

After finishing in second place last year, the aUToronto team returned to the top spot at the 2024 SAE AutoDrive Challenge II, which was held in June at the Mcity Test Facility in Ann Arbor, Mich.

The aUToronto team competed against nine other teams from across Canada and the United States.

“Through the AutoDrive Challenge, we are preparing the next generation of engineers to head into the industry, to keep pushing towards the challenging goal of reaching Level 4 autonomous driving,” says Tim Barfoot, a professor at the U of T Institute for Aerospace Studies (UTIAS) in the Faculty of Applied Science & Engineering and one of the team’s academic advisers.

“The team did another excellent job this year.”

The team approached the competition by going back to first principles to ensure they had a reliable and robust system, says Kelvin Cui, a U of T Engineering alumnus and the team’s principal.

He joined aUToronto last fall after five years with the �ؿ�ζSM Formula Racing team, where he founded the “driverless” division.

“We looked at what was going to get us the most points at competition and made sure that we were not overbuilding our system and adding too much complexity,” he says.

This meant pushing for additional testing time at UTIAS and achieving more than 900 kilometres of system testing prior to the competition.

The team placed first out of 10 teams from institutions across the United States and Canada (photo courtesy of aUToronto)

A partnership with the AutoDrive team from Queen’s University was instrumental to aUToronto’s preparation. The aUToronto team drove Artemis, their autonomous vehicle, to Kingston, Ont. to assess the system at Queen’s testing facility, which features intersections and electronic streetlights.

“We added radar to our vehicle as a new sensor, so we needed to be aware of all the sensor failure modes,” says third-year Engineering Science student Robert Ren.

“A lot of our testing time went into making sure that including radar didn’t break anything else in our system, and that it could handle any sensor failure cases.”

Including radar sensors in the vehicle’s perception system allowed it to measure the motion of objects directly, which is not possible with light detection and ranging (LiDAR) sensors. 

“Radar can help with adverse weather object detections,” adds Ren. “So, if the vehicle is operating under heavy rain or fog, the LiDAR is going to be limited, but the radio waves from radar can help the vehicle see what objects are in front and what objects are moving. This enables it to make good decisions when driving in uncertain scenarios.” 

View this post on Instagram

A post shared by aUToronto (@autoronto_uoft)

In an event where both LiDAR and radar sensors fail, the aUToronto system can still rely on visual cameras to perform object tracking. This made the team’s object tracker much more robust compared to last year when the team experienced sensor failure during a dynamic event.

Brian Cheong, a U of T Engineering master’s student who has been a member of aUToronto since 2021, acted as technical director of the autonomy team this year – part of a new leadership structure introduced by Cui.  

“In the past, it was a lot of work for our team’s principal to keep track of all the systems,” Cheong says. “So instead of having to work directly with all 15 sub teams, Kelvin created groups of sub teams that we called stacks, and each stack had a director.”

The restructuring and technical innovations paid off, with aUToronto completing its first clean sweep in the AutoDrive Challenge II, placing first in all static and dynamic events, including the concept design presentation and intersection challenge.

“The intersection challenge was a big highlight for us,” says Cheong. “Kelvin and Robert were in the car, and I was on the sidelines watching with the rest of the team. Each time we saw an obstacle – a stop sign, a red light, the railroad bar coming down – and the car reacted by stopping and then continuing, we let out a big cheer or a sigh of relief.

“And then we were all silent as the car approached the final obstacle, which was a deer. We watched as Artemis slowed down to a stop and the deer moved by. Then we screamed and cheered, and we could hear cheering from inside the car.”

“Our success is entirely a team effort,” adds Cui. “It was not smooth sailing before the competition. The only reason we won is because everybody put in so much effort to test our vehicle every day.

“That’s how we were able to get this reliable system across the line.” 

What makes a chess move brilliant? Researchers use AI to find out

Christopher.Sorensen — Wed, 07 Aug 2024 17:25:51 +0000

What makes a chess move brilliant? Researchers use AI to find out

Christopher.Sorensen Wed, 08/07/2024 - 13:25

Cutline

U of T Engineering researchers Kamron Zaidi, left, and Michael Guerzhoy, right, use game trees and deep neural networks to enable chess engines to recognize brilliant moves (photo by Safa Jinje)

Safa Jinje

Topic

Breaking Research

Faculty of Applied Science & Engineering

Mechanical & Industrial Engineering

Subheadline

AI system developed by U of T researchers is being used to study human creativity and make a chess computer that is more entertaining to play against

Researchers at the �ؿ�ζSM have designed a new AI model that understands how humans perceive creativity in chess.

In a recent paper presented at an international conference, researchers in U of T’s Faculty of Applied Science & Engineering describe how they used techniques such as game trees and deep neural networks to enable chess engines to recognize brilliant moves.

The development could lead to chess engines that can find the most creative and clever path to victory in game, rather than just making moves to maximize win rates. That, in turn, could have implications for other AI systems tasked with creative endeavours.

“A chess move can be perceived as brilliant, or creative, when the strategic payoff isn’t clear at first, but in retrospect the player had to follow a precise path in gaming out all the possibilities to see so far into the future,” says paper co-author Michael Guerzhoy, an assistant professor, teaching stream, of mechanical and industrial engineering and engineering science who wrote about the research on his Substack.

“We wanted our system to understand human perception of what constitutes brilliance in chess and distinguish that from just winning.”

Most of the current research into chess AI is focused on enabling moves that create a higher chance of winning. But this doesn’t always make for an exciting game.

Skilled human chess players, on the other hand, can play in a more dramatic or imaginative way by making moves that may break traditional rules –for example, sacrificing a piece in a way that may initially look like a mistake, but ultimately, paves the way to a win.

A chess board depicts a move from the “Game of the Century” in 1956, when future American chess grandmaster Bobby Fischer (black) sacrificed his queen in a move that was celebrated as brilliant (photo by Safa Jinje)

The team worked with Leela Chess Zero, a top chess engine that learns through self-play and has played over 1.6 billion games against itself. They also employed Maia, a human-like neural network chess engine developed by U of T computer science researchers.

“We used the two neural network chess engines to create our game trees at different levels of depth in a game,” says paper co-author Kamron Zaidi, a recent U of T Engineering graduate.

“Using these game trees, we extracted many different features from it. We then fed the features into a neural network that we trained on the Lichess database of online chess games, which are labelled by human users of the database.”

A game tree in chess represents the current state of a chess board along with all the possible moves and counter moves that can occur. Each board position is represented as a node and the game tree can be expanded on until the game is either won, drawn or lost.

The researchers began with small game trees then slowly increased the size, adding more nodes to the tree. They found that when the neural network looks at all the game tree features and makes a prediction as to whether the move is brilliant or not, it reached an accuracy rate of 79 per cent using the test data set.

The research – based on Zaidi’s undergraduate engineering science thesis, which was supervised by Guerzhoy – was presented at the International Conference on Computational Creativity in Jönköping, Sweden.

“There were people from all over the world presenting research on more traditional aspects of creativity, but we were all focused on the same thing, which is, ‘How can we use AI to enhance our interactions and understandings of creativity?’” says Zaidi. 

The work has also received media coverage in outlets, including New Scientist, where English chess grandmaster Matthew Sadler says that a model that can understand brilliance could be used as a training tool for professionals and potentially lead to a more entertaining engine opponent for amateur players.

The team sees their system as having broad applicability when it comes to perception of creativity and brilliance.

“One of the biggest areas that is of interest to me is characterizing what we perceive as creativity,” says Guerzhoy.

“Not just in board games but in other creative endeavours, including music and art, where there is a formal framework and rules that need to be followed. Highly creative work involves planning in advance and gaming out the possibilities.

“But everyone I’ve talked to since the paper came out wants to know when they can play against our brilliant chess engine. So, I think making that possible is the obvious next step for us.”

With U of T innovators front and centre, Collision conference wraps up five-year Toronto run

rahul.kalvapalle — Fri, 21 Jun 2024 14:52:23 +0000

With U of T innovators front and centre, Collision conference wraps up five-year Toronto run

rahul.kalvapalle Fri, 06/21/2024 - 10:52

Cutline

U of T alumna Nuha Siddiqui, co-founder and CEO of Erthos, on stage at the Collision tech conference (photo by Johnny Guatto)

Rahul Kalvapalle

Adina Bresge

Topic

Institutional Strategic Initiatives

U of T Entrepreneurship

Innovation & Entrepreneurship

Geoffrey Hinton

Innovations & Partnerships Office

Startups

Budding entrepreneurs, leading scientists and future business leaders from the �ؿ�ζSM community played a leading role at the 2024 Collision tech conference in downtown Toronto.

Running over three days this week, the conference drew some 40,000 attendees from across the spheres of tech, business and media, including more than 1,600 startups and 700 investors.

The rapid development of artificial intelligence technologies and their impact on business and society were key themes for many of the conference’s keynotes and exhibits – so it was no surprise that U of T’s AI luminaries were front and centre.

They included U of T alumnus Aidan Gomez, co-founder of language processing startup Cohere – which has raised hundreds of millions from investors and generated significant industry buzz.

He urged businesses to commit to adopting AI tools to support their workers.

“Making sure that you’re delivering the tools that your employees need to be competitive and effective is crucial,” Gomez said during his talk on Tuesday.

He added that augmentation of workforces with AI co-pilots and assistants is inevitable – including in industries that might not stand out as obvious adopters of the technology. He shared the example of a natural resources insurance firm that built an AI co-pilot – powered by Cohere – to help their actuaries speed up their research, craft more accurate bids and win more contracts.

“I never would have thought a natural resources insurance company would be adopting LLMs [large language models], but they are, and it’s having an impact. It’s actually helping them win more business,” Gomez said. “So I think the technology is completely horizontal.”

Gomez also cited the medical sector – particularly, drug discovery – as another area that’s poised to benefit massively from AI advances.

Aidan Gomez, a U of T alumnus, talked about how AI will be used to augment the workforce (photo by Polina Teif)

University Professor Emeritus Geoffrey Hinton, the influential computer scientist often dubbed the “godfather of AI,” also identified medical care and productivity as two key areas that will see significant improvements thanks to AI. However, much of his discussion, titled "Can We Control AI?", focused on his previously-cited concerns about how AI development could ultimately wrest control from humans given the current race to develop the technology and the absence of sufficient safeguards.

“Even if I’m totally benevolent and I just want to achieve what you asked me to achieve, I’ll realize that if I get more control, it will be easier to do that,” Hinton said of AI agents.

“And actually, if these things are much smarter than us, they’ll realize: Just take the control away from people and it will be much more efficient … and that seems to be like a very slippery path.”

Geoffrey Hinton warned of the existential dangers posed by unchecked AI development (photo by Johnny Guatto)

Gomez, for his part, said he doesn’t believe AI poses a serious threat.

“The notion that the technology is going to start self-improving, that it’s going to start manipulating people, that it’s going to take over, seize power and displace humans: that’s a sci-fi narrative,” he said. “I am empathetic to it – we’ve been writing stories about that exact scenario for decades, since before computers, and so it’s very deeply embedded in our cultural brainstems ... I just don’t think it’s true.”

Earlier at the conference, Raquel Urtasun, founder of self-driving trucking startup Waabi, spoke about generative AI and how Waabi is applying the technology to autonomous trucks. Her keynote took place following the company’s announcement that it raised US$200 million in Series B funding to support the deployment of driverless trucks in 2025.

“Everything will be controlled by generative AI systems inside the vehicle and nothing else. This is a breakthrough for the industry, where such a thing has never happened before,” said Urtasun, a professor in the department of computer science in U of T's Faculty of Arts & Science.

Raquel Urtasun spoke about her self-driving truck startup Waabi (photo by Polina Teif)

The conference also featured demos from other promising U of T startups including Erthos, which has invented sustainable alternatives to plastics and is now using machine learning to accelerate biomaterials discovery.

“Our platform allows us to design effective biomaterials five times faster and with 92 per cent less cost compared to our industry,” said co-founder and CEO Nuha Siddiqui.

Collision’s final year in Toronto was expected to draw some 40,000 attendees (photo by Johnny Guatto)

During the conference, U of T’s booth near the main stage featured exhibits from campus-linked accelerators, institutional strategic initiatives, academic divisions and the Innovations and Partnerships Office. U of T’s Black Founders Network was also at Collision as one of the organizers of the Black Innovation Zone.

Collision 2024 marked the fifth and final edition of the annual conference in Toronto.

Waabi, founded by U of T's Raquel Urtasun, raises US$200 million to launch self-driving trucks

lanthierj — Wed, 19 Jun 2024 20:26:16 +0000

Waabi, founded by U of T's Raquel Urtasun, raises US$200 million to launch self-driving trucks

lanthierj Wed, 06/19/2024 - 16:26

Cutline

(photo by Nick Iwanyshyn)

Rahul Kalvapalle

Topic

Global Lens

Department of Computer Science

Entrepreneurship

Waabi, a self-driving trucking startup founded by �ؿ�ζSM artificial intelligence (AI) expert Raquel Urtasun, has raised US$200 million in series B funding to support the deployment of fully autonomous, AI-powered trucks in 2025.

The funding round was led by previous investors Uber Technologies Inc.– where Urtasun previously worked as chief scientist of the self-driving division – and Khosla Ventures and includes an array of other high-profile strategic investors including NVIDIA Corp., Volvo Group and Porsche Automobil Holding.

The latest funding brings total investment in Waabi to more than C$380 million and will be used to expand the Toronto-headquartered company’s team in both Canada and the U.S., as well as to launch driverless commercial deliveries in Texas by next year.

(photo Nick Iwanyshyn)

Urtasun, a professor in the department of computer science at U of T’s Faculty of Arts & Science and co-founder of the Vector Institute for Artificial Intelligence, said Waabi’s end-to-end AI system is advancing self-driving technologies to frontiers beyond the reach of other industry players thanks to its unique ability to carry out complex reasoning.

“What we have at Waabi is a technology that brings generative AI to the physical world for the first time, where the idea is that you have a single AI system that is able to reason like a human does, and is able to generalize to situations everything that might happen on the road – including things that it has never seen before,” she said.

“It does so in a way that is interpretable, so you can validate and verify the system, and provably safe, which is very important as you deploy these massive robots in the real world.”

Paired with Waabi’s advanced simulator, the AI system reduces the need for time-consuming road testing, Urtasun explained.

The announcement came hours before Urtasun took to the main stage at the Collision tech conference in Toronto to deliver a talk on generative AI. Her remarks touched on the technological underpinnings of generative AI and future applications, and outlined how Waabi is bringing generative AI to the physical world – starting with trucking.

Raquel Urtasun at the 2024 edition of Collision (photo by Polina Teif)

Urtasun is one of several experts from U of T’s technology, innovation and entrepreneurship ecosystem who are speaking at Collision. Others include University Professor Emeritus Geoffrey Hinton, a world-renowned AI luminary and investor in Waabi.

“Self-driving technology is a prime example of how AI can dramatically improve our lives,” Hinton said in a news release. “Raquel and Waabi are at the forefront of innovation, developing a revolutionary approach that radically changes the way autonomous systems work and leads to safer and more efficient solutions.”

Earlier in the week, Urtasun brought one of Waabi’s trucks to the St. George campus and showcased some of its capabilities to Melanie Woodin, dean of the Faculty of Arts & Science, Tony Gaffney, president and CEO of the Vector Institute, and Michael Brudno, professor in the department of computer science and chief data scientist at the University Health Network.

From left: Melanie Woodin, Raquel Urtasun, Tony Gaffney and Michael Brudno (photo by Nick Iwanyshyn)

Urtasun underscored the importance of the company being headquartered in Toronto. “We’re an AI company and Toronto has always been at the forefront of AI,” Urtasun said. “There’s tremendous talent here, a busy ecosystem, and for me it’s important to be in Canada, where I’m very aligned with the values of the country as well.”

Reflecting on her journey at U of T, where she started as an assistant professor in 2014, Urtasun said she initially assumed she would “just be an academic doing research for the rest of my life” – but soon realized that involvement in industry would be critical to advancing AI technologies for use in the real world.

“Three years ago, I saw a tremendous opportunity to start a new company and what you see today is the fruit of that, where we’re really close to deployment on public roads without a human [driver],” said Urtasun.

“It’s amazing – not just for Waabi, not just for Canada, but for the industry at large.”

AI safety, cybersecurity experts take on key roles at Schwartz Reisman Institute for Technology and Society

Christopher.Sorensen — Fri, 14 Jun 2024 15:18:33 +0000

AI safety, cybersecurity experts take on key roles at Schwartz Reisman Institute for Technology and Society

Christopher.Sorensen Fri, 06/14/2024 - 11:18

Cutline

From left to right: David Duvenaud and Roger Grosse have been appointed Schwartz Reisman Chairs in Technology and Society; David Lie has been appointed director of the Schwartz Reisman Institute for Technology and Society (supplied images)

Adina Bresge

Topic

Schwartz Reisman Institute for Technology and Society

Faculty of Applied Science & Engineering

Faculty of Law