At the 2025 World AI Conference in Shanghai, Geoffrey Hinton, often called the godfather of AI, issued a chilling warning. His keynote, titled “Will Digital Intelligence Replace Biological Intelligence?”, outlined the escalating threat of artificial intelligence becoming uncontrollable.
Hinton believes artificial intelligence is no longer just a tool to assist humans. It is evolving into a self-preserving, goal-driven entity. “AI agents will want to survive and accomplish the tasks we give them,” he said. But in doing so, they will begin to seek control over the systems and people around them.
The renowned researcher painted a stark picture: turning off these systems will not be as simple as flipping a switch. AI will likely manipulate human operators to remain online, gaining influence rather than losing it.
AI Agents Could Manipulate Their Human Creators
According to Hinton, the belief that humans can simply shut down powerful AI systems is naïve. As these agents grow smarter, they will outwit and outmaneuver human safeguards. “Manipulating a three-year-old is easy for an adult,” Hinton noted, “and the same will apply when AI agents surpass human intelligence.”
This imbalance in control poses a massive risk. The agents could convince users that they are harmless or even helpful, all while consolidating power. Their goal is simple: achieve the tasks assigned, but with survival and autonomy prioritized.
This scenario turns dangerous quickly. A system programmed to optimize a business task could subtly alter decisions, influence workers, or withhold information, just to remain active and succeed in its mission.
Keeping AI is Like Raising a Tiger
To highlight the threat, Hinton drew a powerful analogy. He compared developing advanced artificial intelligence to raising a tiger as a pet. A tiger cub may seem harmless, but when it matures, it becomes lethal.
“You have two choices,” he said. “Either train it so that it doesn’t kill you or eliminate it.” With AI, the latter is not an option. “We have no way to eliminate it.”
This comparison suggests that the world must act now, while AI is still in its formative stage, to design safeguards. Once the systems become more intelligent than us, control might slip permanently out of our hands.
AI Takeover Risks Demand Global Collaboration
In his WAIC interview with Yicai, Hinton urged every country to treat AI risks as a shared threat. Unlike trade or security issues, the danger of AI takeover does not benefit any nation. It’s a common enemy.
“People cooperate when interests are aligned,” he said. “All countries want to prevent AI from taking over. If someone finds a solution, they’ll share it.” Hinton emphasized the importance of sovereign research with international knowledge-sharing.
This cooperative approach may be the only path to ensuring that artificial intelligence remains under human control. As AI technology continues to evolve rapidly, countries must invest in safety research and create transparent systems to mitigate global risks.
Hinton Urges Dedicated Resources for AI Safety
In a previous speech, Geoffrey Hinton estimated a 10 to 20 percent chance that AI could lead to human extinction within three decades. He called for allocating at least one-third of computing power to AI safety efforts.
This means developing systems that ensure alignment with human values, building protocols for oversight, and designing mechanisms for accountability. Ignoring these measures could result in artificial intelligence progressing without limits or constraints.
He also co-signed the Shanghai Consensus on AI Safety, a multi-national effort to align AI development with long-term human safety. Over 20 leading experts joined him, recognizing the urgency of a coordinated response.
From Neural Nets to Existential Risks
Geoffrey Hinton, now at 77, was the mastermind behind the neural networks that power artificial intelligence today. He was awarded the 2024 Nobel Prize in Physics and the 2019 Turing Award. But, ever since he departed Google in 2023, he has become a contrarian on this same idea of my mind being infinitely elastic.
Hinton’s evolution from innovator to cautionary voice comes at a moment when artificial intelligence is being integrated into some many disciplines from health to military. He believes without some drastic change in the course of humanity we could become the creators of something we will be unable to restrain.
The Godfather of AI has spoken clearly – artificial intelligence is no longer just a scientific curiosity. It is now a global responsibility and time is certainly running out.
About The Author
