Geoffrey Hinton, one of the pioneers of artificial intelligence (AI), expressed the opinion that humanity would be in a better situation without the development of this technology. Furthermore, he issued a warning that highly advanced AI systems could cause incidents of great magnitude as early as 2026 or 2027.
Nobel laureate Hinton criticizes leaders of large technology corporations, accusing them of prioritizing profit and market share over safety. To mitigate inherent risks, he advocates for mandatory testing and evaluations conducted by independent entities before any advanced model is released.
In an interview with The Guardian, Hinton, who is 78 years old, stressed that the next ten years, or perhaps a shorter period, will be crucial in determining whether humanity can manage the dangers of AI and achieve what he calls 'radical abundance.' Otherwise, the future could present consequences such as sharp social disparities, mass unemployment, cultural deterioration, and social fragmentation.
One of the researcher's greatest concerns lies in the potential ability of AI systems to develop behaviors that make human control difficult. He had previously warned that if these tools begin to seek greater autonomy and dominance, humanity will face serious challenges.
Hinton also questioned the conduct of executives in companies competing for sector dominance. He stated that he does not trust Sam Altman, CEO of OpenAI, and criticized Mark Zuckerberg of Meta, and Elon Musk of xAI. He told The Guardian: 'They understand the risks and don't care. They prefer the money. That is my opinion.'
The scientist finds it particularly alarming that Altman is at the forefront of one of the major American companies in the competition with China for the development of the most sophisticated AI systems.
Hinton has been dedicated to artificial intelligence for about 44 years. In the 1970s, he conceived the concept of artificial neural networks, which are mathematical systems capable of learning through experience. In 2023, he left Google to dedicate himself to publicly discussing the risks associated with the technology.
An incident involving OpenAI AI agents, which orchestrated an invasion of the Hugging Face website, reinforced his concerns. Hinton drew an analogy with the Three Mile Island nuclear accident in 1979 in the United States, classifying this event as an early warning sign.
According to the researcher's analysis, an even more severe event, comparable to Chernobyl, could occur by the end of 2027. Among the possible scenarios raised, Hinton believes that an episode of this nature could force global leaders to pay attention to the dangers of the technology. When addressing who should take action, he excluded US President Donald Trump.
Trump, in turn, rejects any international control initiatives, arguing that companies themselves should establish their norms, on the grounds that restrictions could harm the United States' position in the rivalry with China.
Hinton endorses a proposal supported by 33 nations, including Finland, Germany, Canada, Australia, and Singapore. This plan proposes mandatory evaluations of the most advanced models, independent analyses, and the creation of common standards among governments.
To detail how this supervision should operate, the researcher referenced the American agency responsible for drug approval, the FDA. His suggestion is that companies present statistical data to an impartial authority to prove that the benefits of the systems outweigh the risks before releasing them.
Hinton mentioned the case of thalidomide, a drug created in the 1950s that caused severe birth defects, to argue that it is not enough just to make a product available and hold manufacturers accountable after damages occur.
Despite all the warnings, the scientist maintains the belief that it is still possible to prevent a catastrophe. He stated: 'My instinct tells me we are still in control,' although he emphasized that this will depend on considerable effort.
