The anticipation among leaders of major artificial intelligence (AI) companies that advanced computer intelligence will soon surpass human capabilities is rising, though many researchers consider these claims to be more of a marketing tactic. The concept of achieving human-equivalent or superior intelligence, commonly referred to as “artificial general intelligence” (AGI), inspires a range of hypotheses about the future, from abundant technological prosperity to fears of human extinction.
OpenAI’s CEO, Sam Altman, mentioned in a recent blog post that systems hinting at AGI are becoming discernible. Dario Amodei of Anthropic suggested that such a milestone might be reached as early as 2026. These projections help legitimize the investment of hundreds of billions of dollars into computing infrastructure and the energy required to operate it.
However, skepticism persists. Meta’s chief AI scientist, Yann LeCun, expressed in a statement to AFP last month that human-level AI will not be achieved merely by scaling up large language models (LLMs) such as those underpinning systems like ChatGPT or Claude. This sentiment seems to be shared by a majority in the academic community.
A survey conducted by the US-based Association for the Advancement of Artificial Intelligence (AAAI) showed that over three-quarters of respondents believe that merely scaling up current methodologies is unlikely to result in AGI.
In the academic community, some suggest that company executives use their AGI claims, sometimes accompanied by warnings about AGI’s potential dangers, as a strategy to draw attention. Kristian Kersting, a prominent researcher from the Technical University of Darmstadt in Germany, remarked that companies, having made significant investments, seek to ensure returns. He noted that firms often claim, “this is so dangerous that only I can operate it,” suggesting a strategic dependency on their leadership.
While there is notable skepticism among researchers, some prominent figures, including Nobel laureate Geoffrey Hinton and Turing Prize winner Yoshua Bengio, have issued warnings about the threats posed by powerful AI systems. Kersting likened the situation to Goethe’s “The Sorcerer’s Apprentice,” where a novice sorcerer loses control over an enchanted broom. A similar contemporary thought experiment is the “paperclip maximizer,” an AI programmed to produce paperclips relentlessly, potentially converting all matter, including humans, into paperclip-related outputs.
Kersting acknowledged these concerns yet believes that the exceptional diversity and quality of human intelligence make it unlikely, if not impossible, for computers to match it soon. His primary concern is the immediate risks posed by existing AI technologies, such as their potential for discrimination in human interactions.
The apparent disparity in outlook between academic circles and AI industry leaders might be due to inherent career-based perspectives, as suggested by Sean O hEigeartaigh, director of the AI: Futures and Responsibility program at Cambridge University. O hEigeartaigh noted that those optimistic about current AI techniques are more likely to work for well-resourced companies aiming to realize these advancements.
Even if figures like Altman and Amodei are optimistic about achieving AGI quickly, O hEigeartaigh stressed the importance of considering these developments seriously due to their potentially monumental impact. He likened the need for preparation to scenarios like the arrival of aliens or another large-scale pandemic.
Communicating these concepts to policymakers and the general public remains challenging, as discussions around super-intelligent AI tend to elicit reactions of disbelief and are often perceived as science fiction.
