Yoshua Bengio, a tech pioneer, issues a warning about the risks of artificial intelligence turning against humans.

• Yoshua Bengio, a renowned AI pioneer, has cautioned about the potential adverse consequences of AI on society and urged for increased research and safety measures.
• Bengio stated that there are arguments suggesting that the current training methods of AI machines could result in systems that are hostile towards humans.
• "We possess agency to influence the trajectory of societies and humanity positively," he stated.

Yoshua Bengio, a renowned computer scientist and AI pioneer, has cautioned about the dangers of emerging technology on society and urged for more research to minimize its risks.

At the University of Montreal, Bengio, the head of the Montreal Institute for Learning Algorithms, has been awarded for his research in deep learning, a branch of AI that models human brain activity to identify intricate patterns in data.

He expressed concerns about the technology and cautioned that individuals with significant influence might seek to replace humanity with machines.

At the One Young World Summit in Montreal, Bengio told CNBC's Tania Bryer that it is crucial to imagine a future with machines that are as intelligent as us in many ways and to consider the implications for society.

He stated that machines may soon possess the cognitive abilities of humans, with artificial general intelligence (AGI) being a technology that aims to surpass human intellect.

"Power comes from intelligence. But who will control that power?" he asked. "The use of advanced systems can be dangerous if they fall into the wrong hands, leading to instability at a geopolitical level or terrorism."

According to Bengio, only a limited number of organizations and governments will be able to afford to build powerful AI machines, and the bigger the systems are, the smarter they become.

Building and training these machines is extremely expensive, and only a few organizations and countries will be able to do it. This is already the reality, as he stated.

"The concentration of economic, political, and military power poses risks to markets, democracy, and geopolitical stability. As such, we must carefully study these issues and take immediate action to mitigate them."

If it's five years, we won't be ready because we lack methods to ensure that these systems won't harm people or turn against them. We don't know how to do that, he said.

Bengio stated that there are arguments suggesting that the current training methods of AI machines could result in systems that are hostile towards humans.

"While some individuals may seek to exploit this power, others may welcome the demise of humanity in favor of machines. Although a minority, these individuals possess significant influence and can enact their desires unless proper safeguards are implemented immediately," he stated.

AI guidance and regulation

In June, an open letter endorsed by Bengio and signed by current and former employees of Open AI, the company behind ChatGPT, was titled "A right to warn about advanced artificial intelligence."

The advancement of AI poses "serious risks," and OpenAI has been subject to mounting safety concerns, with its "AGI Readiness" team disbanded in October.

"Governments must have regulations that require companies to register when they develop large-scale systems that cost hundreds of millions of dollars to train, according to Bengio. It is crucial for governments to know the specifics of these systems."

Governments must adapt their legislation to technology changes by being creative, as AI is evolving at a rapid pace, according to Bengio.

The computer scientist believes that companies developing AI must be held accountable for their actions.

"Liability can motivate companies to act responsibly, as the threat of lawsuits can push them to prioritize public safety. However, if companies are not liable for their actions, they may not feel the same level of pressure to behave well. This can lead to a race among companies to be the first to achieve AGI, which can be dangerous."

To ensure the safe use of AI, the process of regulating it will be similar to how rules were established for other technologies, such as planes and cars, according to Bengio. "In order to fully harness the potential of AI, we must establish guidelines and oversight to ensure its responsible development," he stated.

Misinformation

The increasing concern about the spread of misinformation, particularly in relation to elections, is exacerbated by the development of AI. In October, OpenAI announced that it had disrupted over 20 operations and deceptive networks worldwide that attempted to use its models. These include social media posts generated by fake accounts ahead of elections in the U.S. and Rwanda.

As we advance towards more advanced systems, the spread of false information and the influence of AI on politics and opinions will become a growing concern, according to Bengio. Meanwhile, the development of increasingly realistic images, voices, and videos will continue to advance, he added.

Bengio stated that the study by Italian and Swiss researchers showed that OpenAI's GPT-4 large language model can persuade people to change their minds better than a human. He added that this influence might extend to interactions with chatbots and could potentially interfere with democratic processes.

The 'hardest question of all'

What does it mean for humanity if we create entities that are smarter than us and have their own goals, as stated by Bengio as the "hardest question of all"?

Bengio stated that the questions are challenging and crucial, yet we lack complete answers. Further research and caution are necessary to minimize the risks.

Bengio urged people to act, stating that "we have agency" and it's not too late to steer the evolution of societies and humanity in a positive and beneficial direction. However, he emphasized that it requires "enough people who understand both the advantages and the risks" and "enough people to work on the solutions." The solutions can be technological, political, or policy-based, but "we need enough effort in those directions right now," Bengio said.

- CNBC's Hayden Field and Sam Shead contributed to this report.