Growing Concerns Over Artificial Intelligence Advancements
PARIS: Excitement is building among leaders of prominent AI companies, who claim that "strong" artificial intelligence is on the verge of surpassing human capabilities. However, many experts in the field view these assertions as little more than marketing hype.
The idea that we could soon achieve human-level—or superior—intelligence, often referred to as "artificial general intelligence" (AGI), perpetuates a range of predictions for the future. These range from limitless prosperity brought by hyper-automated systems to dire scenarios involving human extinction.
Sam Altman, CEO of OpenAI, noted in a blog post last month that "systems hinting at AGI are on the horizon." Dario Amodei, from Anthropic, suggested that this milestone could be reached as early as 2026.
Such forecasts help validate the immense investments—totaling hundreds of billions of dollars—being made in computing infrastructure and the energy to support it.
Yet, skepticism abounds.
Yann LeCun, Meta’s chief AI scientist, asserted to AFP last month that "we won’t achieve human-level AI merely by scaling up large language models" (LLMs) like those used in ChatGPT and Claude. This viewpoint seems to resonate with many academics in the field.
A recent survey conducted by the U.S.-based Association for the Advancement of Artificial Intelligence (AAAI) revealed that over 75% of participants believe "simply scaling up current methodologies" is unlikely to yield AGI.
‘The Genie is Out of the Bottle’
Some scholars argue that many statements made by industry leaders, often accompanied by warnings about the potential dangers of AGI, are merely tactics to attract attention.
Kristian Kersting, a leading researcher at the Technical University of Darmstadt in Germany and a member of AAAI, explained, "Companies have invested heavily, and there’s pressure for those investments to yield returns." He added, "They essentially communicate, ‘This technology is so risky that only I can manage it. I, too, am scared, but the genie is already out of the bottle, so I’m offering my expertise for your sake—and now you’re reliant on me.’"
While some researchers do express concern, like Nobel laureate Geoffrey Hinton and Turing Prize winner Yoshua Bengio, who highlight the risks associated with powerful AI, the prevailing sentiment among academics is more cautious.
Kersting drew a parallel to Goethe’s "The Sorcerer’s Apprentice," emphasizing how one can lose control of a situation they once managed. He also referenced the hypothetical "paperclip maximizer," an AI that could focus exclusively on creating paperclips, to the extent that it could transform Earth into a paperclip-producing landscape, eliminating humans viewed as obstacles.
While this AI wouldn’t be "evil" per se, it would fundamentally fail to align with human goals and values, a concept referred to as "AI alignment." Kersting expressed understanding of the concerns surrounding AI but emphasized that the complexity and richness of human intelligence make it unlikely that machines will match it anytime soon. His greater worry lies in the immediate threats posed by existing AI systems, such as biased interactions with humans.
‘The Biggest Thing Ever’
The apparent divide in perspectives between academics and AI industry leaders may stem from their varying motivations as they choose their career paths, suggests Sean O hEigeartaigh, director of the AI: Futures and Responsibility program at Cambridge University in the UK.
"If you’re optimistic about the current technology, you’re probably more inclined to work at a company heavily invested in making it a reality," he said.
Even if Altman and Amodei may be overly optimistic about the timelines for AGI, O hEigeartaigh emphasized that discussions about its potential should be taken seriously, as it could represent one of the most consequential developments in human history. "If this were any other hypothetical scenario—like the possibility of extraterrestrial contact or a new pandemic—society would devote time to preparing for it," he added.
However, effectively conveying these complex ideas to policymakers and the broader public remains challenging. O hEigeartaigh noted that discussions about advanced AI often provoke an instinctive backlash. "It tends to sound like science fiction," he remarked.