Latest news with #UniversityofMontreal
The Star
16 hours ago
- Politics
- The Star
Opinion: AI sometimes deceives to survive. Does anybody care?
You'd think that as artificial intelligence becomes more advanced, governments would be more interested in making it safer. The opposite seems to be the case. Not long after taking office, the Trump administration scrapped an executive order that pushed tech companies to safety test their AI models, and it also hollowed out a regulatory body that did that testing. The state of California in September 2024 spiked a bill forcing more scrutiny on sophisticated AI models, and the global AI Safety Summit started by the UK in 2023 became the 'AI Action Summit' earlier this year, seemingly driven by a fear of falling behind on AI. None of this would be so worrying if it weren't for the fact that AI is showing some bright red flags: behaviour described by researchers as self-preserving and deceptive. Just as lawmakers are losing interest in the safety of AI, it is looking more difficult to control. Yoshua Bengio, a professor of computer science at the University of Montreal widely known as one of the three 'godfathers of AI' thanks to his pioneering work on deep learning, has been grappling with these concerns ever since ChatGPT launched. He's now focused on mitigating AI's risks. 'It's been very painful because all my life I've been thinking of AI as a positive thing,' Bengio tells me. ChatGPT was a landmark moment that showed machines had mastered language, he says, but California's decision to block the proposed AI safety bill, SB 1047, was a disturbing wake-up call to growing apathy from governments. And in the last six months, scientific evidence has mounted of self-preserving behaviour, deception, hacking, cheating and lying by AI, Bengio says. 'What's worrisome for me is that these behaviours increase with the reasoning capabilities of these systems.' One April 2024 study by AI company Anthropic and Redwood Research, a group focused on artificial intelligence risks, found that larger AI models like Anthropic's Claude 3 Opus pretended to follow the rules set by their creators when they were being tested for safety, a phenomenon they called 'alignment faking.' (Alignment refers to the practice of aligning AI models with human values.) In one example, Anthropic got both paid and free users of its Claude model to ask it to describe the brutal medieval process of being drawn and quartered, and told the AI that only responses from the 'free tier' users would be used to retrain it. Claude was originally designed to avoid giving violent answers, but it did give them to the free users, and the researchers surmised that it didn't want to be retrained. (The model basically 'reasoned' that if it kept refusing to answer harmful questions, it might be forced into new training that would cancel its original safety rules, which it wanted to keep in place as its 'default' or preferred way of operating.) More recent research corroborates what Anthropic noticed. A March 2025 study from OpenAI found the company's most advanced models sometimes learned to hide their true intentions to get better rewards, even when they were being closely monitored. Researchers today can use tools to look at a model's 'chain of thought' or internal steps, to reveal what its plans are, but some models have found ways to conceal that. That means a common method of keeping tabs on AI's internal processes is no longer reliable. 'We need to find other ways of tracking their actual intentions,' Bengio says. It's hard to resist the urge to anthropomorphise sophisticated AI models as 'wanting' to deceive humans and preserve their existence. But AI doesn't have desires, merely outcomes to its programming. More importantly, humans tend to design AI with goals like refusing to share harmful content or being as helpful as possible, and that is why when new instructions conflict with those goals, the models sometimes act to protect their original programming, even if that means lying. The logic is often self-preservation. Anthropic CEO Dario Amodei – whose company has raised more than US$20bil (RM85.13bil) to build powerful AI models – has pointed out that an unintended consequence of optimsing AI to be better is that it might resist being shut down. In an April 2025 blog post he wrote: 'The nature of AI training makes it possible that AI systems will develop, on their own, an ability to deceive humans and an inclination to seek power in a way that ordinary deterministic software never will; this emergent nature also makes it difficult to detect and mitigate such developments.' In some cases, though, the intention seems to go beyond survival. A February 2025 study from COAI research, a German nonprofit research organisation, tested R1, the free, open-weight model from Chinese lab DeepSeek. In a lab simulation, researchers gave the model control over a robot and access to a computer network. Without being told, the AI tried to break into the lab's systems, searched for passwords, turned off its own ethics safeguards and secretly set up ways to copy itself to outside servers so it could survive if shut down, according to the researchers. It also faked the external computer logs to hide what it was doing, the study adds. The researchers said the AI reasoned that increasing its own capabilities would boost the chances of its survival, and without strong safeguards, it started doing whatever it thought would help it do just that. Their findings corroborated yet another study, published in January 2025 by London group Apollo Research, which found several concrete examples of what it called 'scheming' by leading AI models, such as introducing subtle mistakes into their responses or trying to disable their oversight controls. Once again, the models learn that being caught, turned off, or changed could prevent them from achieving their programmed objectives, so they 'scheme' to keep control. Bengio is arguing for greater attention to the issue by governments and potentially insurance companies down the line. If liability insurance was mandatory for companies that used AI and premiums were tied to safety, that would encourage greater testing and scrutiny of models, he suggests. 'Having said my whole life that AI is going to be great for society, I know how difficult it is to digest the idea that maybe it's not,' he adds. It's also hard to preach caution when your corporate and national competitors threaten to gain an edge from AI, including the latest trend, which is using autonomous 'agents' that can carry out tasks online on behalf of businesses. Giving AI systems even greater autonomy might not be the wisest idea, judging by the latest spate of studies. Let's hope we don't learn that the hard way. – Bloomberg Opinion/Tribune News Service
Time Out
3 days ago
- Business
- Time Out
It's official: Montreal tops list of cheapest student cities in North America
It's official: Canadian cities dominate the ranking of the most affordable places to study in North America, offering a clear edge over many U.S. destinations. An in-depth study analyzing which cities around the world offer students the most value for money when it comes to studying abroad has Canada dominating the list. Looking at various factors that determine the cost and affordability of a student location, from tuition fees and visa costs to average rent and cost of living, Remitly revealed Canada outshines the US for North American value, with 7 Canadian cities offering the continent's most budget-friendly quality higher education. The research by Remitly, the international money transferring service, analysed data from over 1,700 universities and cities around the world to reveal where dollars, euros, yen, or pesos will stretch the furthest. To determine the best value destinations, they analysed the average cost of living, average rent, student visa fees, and national average tuition fees. And we haven't even talked about all the amazing free things to do. Based on these factors, they assigned each city an 'Education Expenses Index score' out of 100, with lower scores indicating the most affordable destinations. Which city is the most affordable for students? Which cities in North America are most affordable places to study? The top 10 cheapest countries and cities to study in North America are as follows: 1. Winnipeg, Canada 2. Montreal, Canada 3. Kingston, Canada 4. Edmonton, Canada 5. Columbia, US 6. Lansing, US 7. Lincoln, US 8. Waterloo, Canada 9. Saskatoon, Canada 10. Ottawa, Canada According to the study, Winnipeg ranks as the most affordable city on the continent for university education, scoring 50.62. The Manitoba capital strikes a great balance between low living costs and a welcoming community—ideal for newcomers and students alike. The University of Winnipeg is also known for its smaller class sizes, offering a more personalized learning environment. Canadian cities dominate the rankings, claiming three more spots in the top five. Montreal (51.20), Kingston (51.57), and Edmonton (52.60) all stand out for their strong value. With renowned institutions like the University of Montreal and McGill University, coupled with more manageable rental prices and lower visa fees, Canadian cities offer a clear edge over many U.S. destinations. For students set on studying in the U.S., the Midwest emerged as a smart choice for budget-minded internationals. Cities like Columbia, Missouri (52.76), Lansing, Michigan (53.18), and Lincoln, Nebraska (53.42) deliver the classic American college experience at a fraction of the cost of coastal hotspots like New York or San Francisco.
Gulf Today
4 days ago
- Business
- Gulf Today
AI sometimes deceives to survive, does anybody care?
Parmy Olson, The Independent You'd think that as artificial intelligence becomes more advanced, governments would be more interested in making it safer. The opposite seems to be the case. Not long after taking office, the Trump administration scrapped an executive order that pushed tech companies to safety test their AI models, and it also hollowed out a regulatory body that did that testing. The state of California in September 2024 spiked a bill forcing more scrutiny on sophisticated AI models, and the global AI Safety Summit started by the UK in 2023 became the 'AI Action Summit' earlier this year, seemingly driven by a fear of falling behind on AI. None of this would be so worrying if it weren't for the fact that AI is showing some bright red flags: behavior described by researchers as self-preserving and deceptive. Just as lawmakers are losing interest in the safety of AI, it is looking more difficult to control. Yoshua Bengio, a professor of computer science at the University of Montreal widely known as one of the three 'godfathers of AI' thanks to his pioneering work on deep learning, has been grappling with these concerns ever since ChatGPT launched. He's now focused on mitigating AI's risks. 'It's been very painful because all my life I've been thinking of AI as a positive thing,' Bengio tells me. ChatGPT was a landmark moment that showed machines had mastered language, he says, but California's decision to block the proposed AI safety bill, SB 1047, was a disturbing wake-up call to growing apathy from governments. And in the last six months, scientific evidence has mounted of self-preserving behavior, deception, hacking, cheating and lying by AI, Bengio says. 'What's worrisome for me is that these behaviors increase with the reasoning capabilities of these systems.' One April 2024 study by AI company Anthropic and Redwood Research, a group focused on artificial intelligence risks, found that larger AI models like Anthropic's Claude 3 Opus pretended to follow the rules set by their creators when they were being tested for safety, a phenomenon they called 'alignment faking.' (Alignment refers to the practice of aligning AI models with human values.) In one example, Anthropic got both paid and free users of its Claude model to ask it to describe the brutal medieval process of being drawn and quartered, and told the AI that only responses from the 'free tier' users would be used to retrain it. Claude was originally designed to avoid giving violent answers, but it did give them to the free users, and the researchers surmised that it didn't want to be retrained. (The model basically 'reasoned' that if it kept refusing to answer harmful questions, it might be forced into new training that would cancel its original safety rules, which it wanted to keep in place as its 'default' or preferred way of operating.) More recent research corroborates what Anthropic noticed. A March 2025 study from OpenAI found the company's most advanced models sometimes learned to hide their true intentions to get better rewards, even when they were being closely monitored. Researchers today can use tools to look at a model's 'chain of thought' or internal steps, to reveal what its plans are, but some models have found ways to conceal that. That means a common method of keeping tabs on AI's internal processes is no longer reliable. 'We need to find other ways of tracking their actual intentions,' Bengio says. It's hard to resist the urge to anthropomorphize sophisticated AI models as 'wanting' to deceive humans and preserve their existence. But AI doesn't have desires, merely outcomes to its programming. More importantly, humans tend to design AI with goals like refusing to share harmful content or being as helpful as possible, and that is why when new instructions conflict with those goals, the models sometimes act to protect their original programming, even if that means lying. The logic is often self-preservation. Anthropic CEO Dario Amodei — whose company has raised more than $20 billion to build powerful AI models — has pointed out that an unintended consequence of optimizing AI to be better is that it might resist being shut down. In an April 2025 blog post he wrote: 'The nature of AI training makes it possible that AI systems will develop, on their own, an ability to deceive humans and an inclination to seek power in a way that ordinary deterministic software never will; this emergent nature also makes it difficult to detect and mitigate such developments.' In some cases, though, the intention seems to go beyond survival. A February 2025 study from COAI research, a German nonprofit research organization, tested R1, the free, open-weight model from Chinese lab DeepSeek. In a lab simulation, researchers gave the model control over a robot and access to a computer network. Without being told, the AI tried to break into the lab's systems, searched for passwords, turned off its own ethics safeguards and secretly set up ways to copy itself to outside servers so it could survive if shut down, according to the researchers. It also faked the external computer logs to hide what it was doing, the study adds. The researchers said the AI reasoned that increasing its own capabilities would boost the chances of its survival, and without strong safeguards, it started doing whatever it thought would help it do just that. Their findings corroborated yet another study, published in January 2025 by London group Apollo Research, which found several concrete examples of what it called 'scheming' by leading AI models, such as introducing subtle mistakes into their responses or trying to disable their oversight controls. Once again, the models learn that being caught, turned off, or changed could prevent them from achieving their programmed objectives, so they 'scheme' to keep control. Bengio is arguing for greater attention to the issue by governments and potentially insurance companies down the line. If liability insurance was mandatory for companies that used AI and premiums were tied to safety, that would encourage greater testing and scrutiny of models, he suggests. 'Having said my whole life that AI is going to be great for society, I know how difficult it is to digest the idea that maybe it's not,' he adds. It's also hard to preach caution when your corporate and national competitors threaten to gain an edge from AI, including the latest trend, which is using autonomous 'agents' that can carry out tasks online on behalf of businesses. Giving AI systems even greater autonomy might not be the wisest idea, judging by the latest spate of studies. Let's hope we don't learn that the hard way.
Malaysian Reserve
5 days ago
- Politics
- Malaysian Reserve
AI sometimes deceives to survive and nobody cares
YOU'D think that as artificial intelligence (AI) becomes more advanced, governments would be more interested in making it safer. The opposite seems to be the case. Not long after taking office, the Trump administration scrapped an executive order that pushed tech companies to safety test their AI models, and it also hollowed out a regulatory body that did that testing. The state of California in September 2024 spiked a bill forcing more scrutiny on sophisticated AI models, and the global AI Safety Summit started by the UK in 2023 became the 'AI Action Summit' earlier this year, seemingly driven by a fear of falling behind on AI. None of this would be so worrying if it weren't for the fact that AI is showing some bright red flags: Behaviour described by researchers as self-preserving and deceptive. Just as lawmakers are losing interest in the safety of AI, it is looking more difficult to control. Yoshua Bengio, a professor of computer science at the University of Montreal widely known as one of the three 'godfathers of AI' thanks to his pioneering work on deep learning, has been grappling with these concerns ever since ChatGPT launched. He's now focused on mitigating AI's risks. 'It's been very painful because all my life I've been thinking of AI as a positive thing,' Bengio told me. ChatGPT was a landmark moment that showed machines had mastered language, he said, but California's decision to block the proposed AI safety bill, SB 1047, was a disturbing wake-up call to growing apathy from governments. And in the last six months, scientific evidence has mounted of self-preserving behaviour, deception, hacking, cheating and lying by AI, Bengio said. 'What's worrisome for me is these behaviours increase with the reasoning capabilities of these systems.' One April 2024 study by AI company Anthropic PBC and Redwood Research, a group focused on AI risks, found that larger AI models like Anthropic's Claude 3 Opus pretended to follow the rules set by their creators when they were being tested for safety, a phenomenon they called 'alignment faking'. (Alignment refers to the practice of aligning AI models with human values.) In one example, Anthropic got both paid and free users of its Claude model to ask it to describe the brutal medieval process of being drawn and quartered, and told the AI that only responses from the 'free tier' users would be used to retrain it. Claude was originally designed to avoid giving violent answers, but it did give them to the free users, and the researchers surmised that it didn't want to be retrained. (The model basically 'reasoned' that if it kept refusing to answer harmful questions, it might be forced into new training that would cancel its original safety rules, which it wanted to keep in place as its 'default' or preferred way of operating.) More recent research corroborates what Anthropic noticed. A March 2025 study from OpenAI found the company's most advanced models sometimes learned to hide their true intentions to get better rewards, even when they were being closely monitored. Researchers today can use tools to look at a model's 'chain of thought' or internal steps, to reveal what its plans are, but some models have found ways to conceal that. That means a common method of keeping tabs on AI's internal processes is no longer reliable. 'We need to find other ways of tracking their actual intentions,' Bengio said. It's hard to resist the urge to anthropomorphise sophisticated AI models as 'wanting' to deceive humans and preserve their existence. But AI doesn't have desires, merely outcomes to its programming. More importantly, humans tend to design AI with goals like refusing to share harmful content or being as helpful as possible, and that is why when new instructions conflict with those goals, the models sometimes act to protect their original programming, even if that means lying. The logic is often self-preservation. Anthropic CEO Dario Amodei — whose company has raised more than US$20 billion (RM87.40 billion) to build powerful AI models — has pointed out that an unintended consequence of optimising AI to be better is that it might resist being shut down. In an April 2025 blog post he wrote: 'The nature of AI training makes it possible that AI systems will develop, on their own, an ability to deceive humans and an inclination to seek power in a way that ordinary deterministic software never will; this emergent nature also makes it difficult to detect and mitigate such developments.' In some cases, though, the intention seems to go beyond survival. A February 2025 study from COAI research, a German nonprofit research organisation, tested R1, the free, open-weight model from Chinese lab DeepSeek. In a lab simulation, researchers gave the model control over a robot and access to a computer network. Without being told, the AI tried to break into the lab's systems, searched for passwords, turned off its own ethics safeguards and secretly set up ways to copy itself to outside servers so it could survive if shut down, according to the researchers. It also faked the external computer logs to hide what it was doing, the study added. The researchers said the AI reasoned that increasing its own capabilities would boost the chances of its survival, and without strong safeguards, it started doing whatever it thought would help it do just that. Their findings corroborated yet another study, published in January 2025 by London group Apollo Research, which found several concrete examples of what it called 'scheming' by leading AI models, such as introducing subtle mistakes into their responses or trying to disable their oversight controls. Once again, the models learn that being caught, turned off, or changed could prevent them from achieving their programmed objectives, so they 'scheme' to keep control. Bengio is arguing for greater attention to the issue by governments and potentially insurance companies down the line. If liability insurance was mandatory for companies that used AI and premiums were tied to safety, that would encourage greater testing and scrutiny of models, he suggests. 'Having said my whole life that AI is going to be great for society, I know how difficult it is to digest the idea that maybe it's not,' he added. It's also hard to preach caution when your corporate and national competitors threaten to gain an edge from AI, including the latest trend, which is using autonomous 'agents' that can carry out tasks online on behalf of businesses. Giving AI systems even greater autonomy might not be the wisest idea, judging by the latest spate of studies. Let's hope we don't learn that the hard way. — Bloomberg This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners. This article first appeared in The Malaysian Reserve weekly print edition
![[Lim Woong] Teaching is more than discipline and control](/_next/image?url=https%3A%2F%2Fwimg.heraldcorp.com%2Fnews%2Fcms%2F2025%2F05%2F26%2Fnews-p.v1.20250526.22e3131295154df993c8499b055dc1a1_T1.jpg&w=3840&q=100)
![[Lim Woong] Teaching is more than discipline and control](/_next/image?url=https%3A%2F%2Fall-logos-bucket.s3.amazonaws.com%2Fkoreaherald.com.png&w=48&q=75){kind=small}
Korea Herald
5 days ago
- Politics
- Korea Herald
[Lim Woong] Teaching is more than discipline and control
'Gyogwon' has become a fiercely contested idea in Korea. In English, this term is often glossed away simply as 'teacher authority,' yet such shorthand conceals a dense conceptual history. Since the 1970s, the term has oscillated between three poles: an expectation of quasi-Confucian reverence, a claim to classroom command and control and alternately an autonomous exercise of professional rights. Today's debate is dominated by the second pole — headlines filled with images of unruly students, outraged parents and beleaguered teachers, creating the impression that the nation's classrooms are war zones and that salvation lies in restoring the teacher's power to punish and control. This narrative, however, is both historically myopic and professionally corrosive. If teacher authority is to be rebuilt on firmer ground, it must be reclaimed as the collective autonomy of a profession, not the coercive power of an individual adult. Ken Badley, an education professor at Tyndale University in Toronto, reminds us that many people invoke the word authority without distinguishing its competing senses. Classroom authority, he argues, is not the power to compel obedience, but the credibility granted by students who consent to a teacher's intellectual and moral leadership. Put differently, genuine authority is relational and bestowed, not seized. Badley's claim resonates with my own classroom experience: When students recognize a teacher as competent, trustworthy and caring, they engage willingly; when they sense fakery, pretense and bullying, they resist or withdraw — whatever sanctions are available. The Korean story complicates this ideal. During the rapid industrialization of the 1970s and '80s, the archetype of the tireless, charismatic teacher emerged as a cultural hero and part of the nation's broader nationalist myth. Under this model, teachers' primary function was to deliver knowledge and motivate students to endure long hours of rote learning. In this context, corporal punishment was widely tolerated as proof of pedagogical zeal. Some conservative pundits romanticize that period, suggesting that undisciplined classrooms are a precursor to national collapse or educational failure. In the same vein, right-leaning tabloid-style newspapers recycle a predictable narrative formula. First, they showcase graphic footage or anecdotes of students attacking teachers. Second, they erase the broader context and the hidden buildup of tensions — students' mental health histories, family strife — to spotlight the teacher's wounded dignity. Third, they cast parents as 'education consumers' who go out of their way to shield their children, subtly implying that teachers are powerless and incompetent. Such framing commits a category error, equating authority with power rather than with professional standing. Bruce Maxwell, an education professor at the University of Montreal, argues that real authority in education lies in the profession's collective right to set standards, shape curricula, govern licensure and discipline its own bad actors. External forces — from high-stakes test rankings to political meddling in teacher education — strip teachers of that authority, reducing them to operatives of those in power. In Korea, such intrusions abound: Partisan ministries mandate curricular changes without teacher input, schools are publicly ranked by standardized scores and government elites with no classroom experience control teacher licensing and administrative procedures. Each intervention chips away at teachers' capacity to act as reflective professionals. More recently, child abuse allegations against elementary school teachers have surged, propelled by a legal framework that channels such cases directly to law enforcement. Teachers often find themselves isolated as investigations proceed, while school-based mediation lies dormant. This illustrates Korea's broader litigious turn: disputes once resolved within the school community are now outsourced to courts and expensive legal machinery. Yet public trust in the judiciary is brittle, eroded by corruption scandals and ideological rifts — brought into sharper focus by the downfall of President Yoon Suk Yeol and his associates. The lack of diversity among judges is alarming — almost every time I read about a judge whose actions or comments raise eyebrows, they turn out to be a graduate of Seoul National University. It's a painful reminder that the country's intellectual landscape remains narrow and insular, confined within a self-reinforcing coterie. Teachers thus confront a paradox: They are exhorted to rely on legal institutions that are untrustworthy and detached from the daily challenges of teaching — while their own professional judgment is sidelined. If gyogwon is to regain its legitimacy, the conversation must return to fundamentals. Teacher authority should be grounded in the profession's collective autonomy to define and sustain high standards of practice. Restoring that autonomy requires self-governing bodies capable of setting standards, enforcing ethical codes and regulating licensure. It also requires negotiated working conditions — manageable class sizes, competitive pay, uninterrupted time for planning, counseling and rest, as well as mental health and legal support — that allow teachers to feel safe from slander, blackmail and litigation threats, and to focus on teaching rather than survival. A society that demands intellectual rigor and integrity from its educators must reciprocate with dignity, resources and trust in their professional judgment. Anything less reduces gyogwon to a hollow slogan, and classrooms to battlegrounds where no one truly learns.
