An arms race for artificial intelligence (AI) supremacy, sparked by recent concerns over the Chinese chatbot DeepSeek, poses a risk of escalating the existential threats of superintelligence, as cautioned by one of the pioneers in the field of AI.
Yoshua Bengio, a Canadian machine learning pioneer and author of the forthcoming International AI Safety Report, emphasizes the importance of oversight in the unchecked pursuit of computational power for AI.
“The focus is on winning the race, rather than ensuring that we do not create something that could backfire,” Mr. Bengio explains.
He warns against the potential consequences of military and economic competitions, such as compromising ethics, responsibility, and safety, which are inevitable outcomes when the race for AI supremacy takes precedence.
Mr. Bengio, known for his work on neural networks and machine learning, is currently in London to receive the prestigious Queen Elizabeth Prize for Engineering, recognizing AI’s transformative potential.
Despite acknowledging the positive societal impacts of AI, he expresses concerns about the shift away from AI regulation and the intense competition among tech giants for more powerful AI models.
‘Superhuman systems becoming more powerful’
“We are creating increasingly more potent systems that are surpassing human capabilities in certain aspects,” he notes.
“As these systems gain more power, they also become significantly more valuable economically.
“The desire for substantial financial gain motivates many people. Naturally, when selling products, discussing the risks is not a priority.”
Not all AI experts share the same level of concern, however.
Yann LeCun, Meta’s chief AI scientist, attending the QEPrize ceremony in London, argues that the current large language models are not truly intelligent.
“We do not yet have machines that possess the level of intelligence equivalent to that of a house cat when it comes to understanding the physical world,” he explains.
Mr. LeCun predicts that within the next three to five years, AI will exhibit some aspects of human-level intelligence, enabling robots to perform tasks beyond their programmed capabilities.
Read more:
What is DeepSeek? The low-cost Chinese AI firm that has turned the tech world upside down
Bill Gates says he would be diagnosed with autism if he was young today
Contrary to concerns about AI dominance, Mr. LeCun points to open-source AI models like DeepSeek, developed by a Chinese company, showcasing innovation that disrupts the notion of long-term dominance by any single entity.
“If the US chooses to limit AI development for geopolitical or commercial reasons, innovation will emerge elsewhere in the world. DeepSeek’s success highlights this,” he suggests.
The Queen Elizabeth Prize for Engineering, awarded annually to engineers making significant global impacts, focuses on revolutionary discoveries.
Previous recipients include innovators in photovoltaic cells, wind turbine technology, neodymium magnets, and electric motors.
Lord Vallance, the science minister and chairman of the QEPrize foundation, acknowledges the potential risks associated with AI.
Organizations like the UK’s AI Safety Institute aim to anticipate and mitigate the potential dangers posed by AI advancing to human-like intelligence.
Science minister Lord Vallance
However, he remains unconcerned about the prospect of a single nation or company monopolizing AI.
“Recent developments suggest that multiple companies will play a significant role in the AI space, making the idea of one entity dominating the field less probable,” he concludes.