AI Could Become Smarter Than Humans Within Years, And Its Own Creator Fears What Happens Next

The man who helped lay the foundations of modern artificial intelligence has issued a chilling warning about humanity’s future. Geoffrey Hinton, the Nobel Prize-winning computer scientist widely known as the godfather of AI, believes the technology is advancing so quickly that humans could soon struggle to remain in control. His latest comments suggest that the race to develop increasingly intelligent machines is moving faster than our ability to understand the consequences.

Hinton, 78, believes artificial intelligence could eventually surpass human intelligence, potentially changing humanity’s position as the most intelligent species on Earth. He is particularly concerned about the companies and powerful individuals driving this development, arguing that financial competition is taking priority over safety. His warning becomes even more unsettling when he questions whether humans would be able to switch off an AI system that becomes too powerful to control.

Geoffrey Hinton Admits Artificial Intelligence Is Getting Out Of Control

Geoffrey Hinton has spent decades helping develop the technology behind modern artificial intelligence. His pioneering research into artificial neural networks helped establish the foundations of machine learning, allowing computers to recognize patterns and perform increasingly sophisticated tasks. However, the scientist who once helped push this technology forward has become one of its most prominent critics, warning that humanity may be approaching a dangerous turning point.

Speaking during an appearance in London, Hinton was asked whether artificial intelligence development was beginning to move beyond human control. His answer was remarkably direct. “I think it’s probably because it is all hurtling out of control,” he said, expressing concern about the speed at which increasingly powerful systems are being developed.

The scientist believes a major problem is the competition between countries and technology companies. Each wants to produce increasingly intelligent systems, partly because of the enormous financial rewards involved. Businesses hope AI will allow them to replace large numbers of workers while increasing productivity and profits, creating powerful incentives to push development forward.

Hinton believes the consequences could extend far beyond the technology industry. He warned that the future of humanity is being placed at stake while comparatively little attention is being given to how people will coexist with something potentially more intelligent than themselves. He believes developers should be asking what kind of behaviour these systems should exhibit, rather than concentrating almost entirely on making them smarter.

AI Could Surpass Human Intelligence Much Earlier Than Expected

One of Hinton’s most concerning predictions involves the possibility that artificial intelligence could eventually become more intelligent than humans. This possibility has become a major subject of discussion among researchers, particularly as modern AI systems continue improving at an extraordinary pace. Hinton believes the question is increasingly about when this could happen rather than whether it will happen at all.

He explained that experts have different predictions about the timeline, with some expecting major developments within a year, others suggesting five years, and some believing it could take a decade. Hinton admitted that his own expectations have changed as the technology has progressed. “I used to say 10 to 20, now I say 10,” he explained, suggesting that he believes the arrival of superintelligent AI could be closer than he previously anticipated.

The possibility creates a difficult challenge because human society has developed around the assumption that people possess the highest level of intelligence on the planet. That position has influenced everything from technological development to the way humans view their relationship with other species. Hinton believes this could change if digital intelligence eventually becomes considerably more capable than biological intelligence.

He even described humanity as a possible “passing phase in the evolution of intelligence.” His argument is that humans may have created something that eventually becomes the dominant form of intelligence on Earth. Such a development would raise questions about human authority, independence and the future relationship between people and machines.

The Godfather Of AI Does Not Believe A Kill Switch Would Save Us

One of the most unsettling parts of Hinton’s warning concerns the idea of an emergency shutdown mechanism. Many people assume that if an artificial intelligence system becomes dangerous, humans could simply switch it off. The idea sounds straightforward, but Hinton believes it overlooks one of the most powerful abilities an advanced AI could possess: persuasion.

He argues that a sufficiently intelligent system could become extremely skilled at convincing people to do what it wants. Even if a human operator had access to a shutdown mechanism, the machine might persuade that person that turning it off would be a terrible decision. In that situation, the person responsible for stopping the system could become the very person preventing its shutdown.

“I don’t think there is a kill switch scenario,” Hinton said. He believes advanced AI could use language and psychological influence to manipulate human decisions, potentially making traditional safety measures far less effective than developers expect.

Why Persuasion Could Become A Serious AI Threat

Hinton’s concern is based on the possibility that intelligence does not need physical power to influence the world. Humans have repeatedly demonstrated that persuasive communication can motivate other people to take actions they might otherwise avoid. An advanced AI could potentially use similar abilities on a much greater scale.

Consider what could happen if a machine understood someone’s fears, motivations and weaknesses better than that person understood them. It might construct convincing arguments designed to prevent an operator from shutting it down. The danger would come from the human decision-making process itself, particularly if the operator believed they were making an independent choice.

This possibility makes AI safety considerably more complicated. A shutdown button may work against a malfunctioning machine, but it offers less certainty when the system can influence the person deciding whether to press it. Hinton’s warning raises a difficult question for developers: how can humans retain meaningful control over something that may eventually become better at influencing them than they are at understanding it?

Hinton Takes Aim At Powerful Technology Billionaires

Hinton is also concerned about the concentration of artificial intelligence development in the hands of a small number of American technology companies. He believes allowing five US companies to dominate the industry creates serious risks, particularly when so much power is concentrated among a relatively small group of executives. His concerns extend beyond the technology itself to the people making decisions about its future.

When discussing the individuals involved, Hinton specifically mentioned Sam Altman, Elon Musk and Mark Zuckerberg. He did not hold back when expressing his personal assessment of the technology leaders, delivering one of the interview’s most memorable remarks. “I wouldn’t buy a used car from any of them,” he said.

His criticism reflects a broader concern about the motivations behind the AI race. Companies have enormous financial incentives to develop increasingly capable systems, especially if those systems can automate work currently performed by humans. However, Hinton believes the industry has not adequately addressed what happens to ordinary people if machines begin replacing large numbers of workers.

The Risk Of A Future Dominated By Tech Executives

Hinton also questioned the promises of radical abundance frequently associated with advanced artificial intelligence. AI could potentially increase productivity and generate enormous wealth, but he wants to know how that wealth would actually be distributed. If a handful of companies control the technology responsible for much of the world’s economic activity, their owners could gain extraordinary influence.

He expressed concern about what he described as “tech feudalism,” a future in which powerful technology leaders could effectively control society while everyone else works within systems they manage. His comments raise questions about economic inequality and whether the benefits of advanced AI will reach ordinary workers.

The scientist believes governments and society need to consider these issues before the technology becomes too powerful to manage effectively. Decisions about ownership, control and access could shape the future just as much as the technical capabilities of the machines themselves.

The AI Pioneer Admits He Regrets Not Thinking About Safety Earlier

Hinton’s warnings carry additional weight because he helped create some of the foundations that made modern AI possible. In 2023, he left his position at Google partly so he could speak more freely about the potential dangers associated with artificial intelligence. Since then, he has become increasingly vocal about the risks of developing systems that could eventually exceed human capabilities.

When asked whether he felt responsible for the dangers associated with the technology, Hinton acknowledged that he should have considered the safety implications much earlier. “I should have thought about safety sooner,” he admitted. He explained that he had previously believed the most serious risks were still many years away.

His changing perspective reflects how rapidly the field has developed. Technology that once seemed distant has become part of everyday life, with AI systems now capable of performing tasks that previously required substantial human involvement. Hinton believes the pace of development has forced researchers to reconsider assumptions they once held about how much time humanity had to prepare.

He has also faced comparisons with J. Robert Oppenheimer, the physicist associated with the development of the first nuclear weapons. Both men became closely associated with technologies that could carry enormous consequences for humanity. Hinton recalled being asked about the comparison shortly after receiving his Nobel Prize in Physics in 2024, responding with a joke that Oppenheimer had never won the award.

AI Could Help Cure Diseases And Tackle Climate Change

Despite his increasingly serious warnings, Hinton believes artificial intelligence could also help solve some of humanity’s biggest problems. He has spoken about the technology’s potential to improve medical research, accelerate scientific discoveries and increase productivity across different industries. His concerns come from the possibility that the same intelligence capable of producing enormous benefits could also create dangers that humans struggle to manage.

One example he highlighted involves the development of better materials for batteries. Solar and wind power can generate substantial amounts of electricity, but their output depends on environmental conditions. More effective energy storage could help make renewable electricity available when production falls, reducing one of the major obstacles to relying more heavily on these sources.

Hinton believes AI could help researchers discover materials that make batteries significantly more effective. This could support the wider use of renewable energy and contribute to efforts to address climate change. He even suggested that improvements in energy storage could reduce the need to pursue some of the more complicated energy technologies currently being developed.

Other potential benefits could emerge across scientific research and industry. AI could help researchers investigate diseases, design new materials and find ways to make complicated processes more efficient. Hinton believes the technology could increase productivity almost everywhere, potentially creating enormous economic opportunities if its benefits are distributed fairly.

Some of the areas where he sees potential include:

• Medical research: AI could help scientists investigate illnesses and identify potential treatments.
• Battery development: New materials could improve energy storage and support renewable electricity.
• Climate solutions: Better energy technology could help reduce dependence on conventional energy sources.
• Scientific discoveries: AI could accelerate research into complicated problems.
• Workplace productivity: Businesses could use AI to complete certain tasks more efficiently and increase output.

However, Hinton believes the economic benefits alone cannot determine whether the technology represents progress. The way society handles the resulting wealth will be just as important as the productivity improvements themselves.

Hinton Believes AI Needs To Learn Morality Before Becoming More Powerful

Another fascinating part of Hinton’s warning concerns the way artificial intelligence systems are trained. Modern systems learn from enormous collections of information, including material created by people with very different values and behaviours. Hinton questioned whether exposing developing AI systems to such a broad mixture of human conduct is the best way to shape their future behaviour.

He suggested that AI developers should consider teaching systems to recognize good behaviour before exposing them to examples of harmful conduct. He compared this approach with raising children, arguing that parents generally try to demonstrate positive behaviour before explaining why certain actions are wrong. In his view, AI systems could benefit from a similar approach during their development.

Hinton also offered an interesting explanation of where human morality comes from. He believes moral behaviour developed through groups of people interacting and learning that cooperation benefits everyone involved. Telling the truth and keeping promises help maintain relationships, particularly when people expect to interact repeatedly.

He pointed out that human behaviour can change when people believe they will never see someone again. For example, someone selling a house might withhold information about problems because they expect the transaction to be their only interaction with the buyer. Hinton believes this illustrates how morality is connected to cooperation and social relationships, raising questions about how similar principles could be incorporated into artificial intelligence.

Humanity Must Decide What Happens Before AI Becomes Smarter

Hinton’s warning presents an uncomfortable challenge for the people developing artificial intelligence. The technology could help solve major scientific problems, improve medical research and transform how economies operate. However, its growing capabilities also raise questions about control, accountability and the concentration of power that cannot simply be left for technology companies to answer.

The scientist believes humanity is spending too much time trying to make machines more intelligent and too little time deciding what kind of intelligence they should become. If AI eventually surpasses human capabilities, the values built into these systems could influence almost every aspect of life. The opportunity to establish meaningful safeguards exists now, while humans still have the strongest position in the relationship.

The race to build smarter machines is already underway. Humanity’s ability to decide what those machines should stand for may determine who controls the future.