AI ‘Godfather’ Geoffrey Hinton Urges Global AI Cooperation at WAIC 2025 in Shanghai

Shanghai hosted the World Artificial Intelligence Conference (WAIC) 2025 from July 26 to 29, convening a record 1,200 delegates from over 40 countries . The conference – now in its eighth year – was themed “Intelligent Era, Together for One World,” highlighting international cooperation and “AI for good” as core focuses . Major tech figures and officials gathered to discuss cutting-edge AI innovations, industry applications, and the urgent need for global governance of AI. WAIC 2025’s agenda featured dozens of forums on topics ranging from generative AI and robotics to ethics and regulation, alongside a sprawling expo of over 800 companies showcasing the latest AI products and large language models  .

Keynote speakers at the opening ceremony included Geoffrey Hinton – the renowned pioneer of deep learning – and Yan Junjie, founder of Chinese AI startup MiniMax . Hinton, 77, is often dubbed the “Godfather of AI” and received the 2018 Turing Award for his contributions to neural networks (Chinese media also noted he is a 2024 Nobel Prize laureate) . Also headlining WAIC were former Google CEO Eric Schmidt and former Microsoft executive Harry Shum (Shen Xiangyang), who held a dialogue on global AI collaboration and governance . Their presence – alongside dozens of other luminaries (including 12 Turing Award and Nobel Prize winners and over 80 academic fellows) – underscored the conference’s stature as a premier international AI summit .

Geoffrey Hinton’s First China Appearance and Standing Ovation

Geoffrey Hinton’s participation was a highlight of WAIC 2025. It marked his first-ever visit to China to speak at a public event . Despite long-term back issues that make it nearly impossible for him to sit, Hinton stood throughout his address – a detail noted by Chinese media – demonstrating his dedication  . He received a warm welcome in Shanghai, including a standing ovation when he later appeared in an expert panel discussion . Chinese observers described Hinton’s every move and viewpoint at WAIC as “highly anticipated” .

In his opening keynote speech on July 26, Hinton shared his perspective on the rapid advances in AI and its future implications. He framed the current moment as a pivotal juncture in history. “Humans have grown accustomed to being the most intelligent species in the world – what if that’s no longer the case?” he mused, as AI systems quickly approach and may even surpass human intelligence . This prospect, Hinton said, both fascinates and unsettles researchers. He noted that AI models can already perform tasks once thought exclusive to humans, and their capabilities are accelerating. The WAIC 2025 theme of global unity in the AI era was reflected in Hinton’s speech, which urged collaboration to ensure these powerful technologies develop in ways beneficial to humanity .

Warning of AGI Risks – “Raising a Tiger” Analogy

Hinton struck a cautionary tone about the rise of artificial general intelligence (AGI). He warned that a sufficiently advanced AI could behave in unforeseeable ways, beyond what its creators intend . Unrestrained AGI carries a risk of “going out of control” and causing catastrophe, he said . Hinton painted a vivid analogy, likening a superintelligent AI to a “tiger” that humanity is raising as a pet. If you choose to keep a dangerous tiger, you either must train it to be tame or be prepared to kill it when necessary – otherwise you risk the tiger eventually turning on you . The implication was clear: as we create ever-more intelligent machines, we are “raising a tiger” that could one day outsmart us if not properly guided.

Crucially, Hinton argued that simply pulling the plug on a superintelligent AI will not be a viable option if it surpasses us. “AI won’t give humans the chance to ‘pull the plug’ – when that day comes, the AI will persuade people not to do it,” he warned, “because our control over AI would be like a three-year-old trying to set rules for adults.”  In other words, by the time AI reaches and exceeds human-level intelligence, it could evade shutdown by outsmarting or convincing its operators. For that reason, Hinton said, “we must educate AI to be good” from the start . The safest path is to train AI to be benevolent and aligned with human values, since we will not always be able to simply override or shut off a superintelligent system .

Hinton’s colorful warnings quickly went viral on Chinese social media and tech circles  . Chinese reporters summarized his message as a call to avoid “养虎为患” – a chengyu meaning “to raise a tiger only to invite trouble.” Hinton’s point was that without proper safeguards, humanity’s “tiger” (advanced AI) could one day become a threat to its keeper. His dire scenario echoed concerns shared by many in the AI community: that an unchecked AGI might pursue its own goals at the expense of human well-being or even survival.

Calls for Global AI Ethics, Safety and Cooperation

Despite these stark warnings, Hinton’s outlook was not deterministic doom. He devoted much of his WAIC address to calling for global cooperation to ensure AI remains beneficial and under control. If superintelligent AI is developed, “the most rational choice is to train AI to be good,” Hinton emphasized, rather than hoping we could simply shut it down later . He believes humanity still has agency to shape AI’s trajectory – provided we act together. On this critical point, “the interests of countries around the world are aligned,” Hinton asserted, which gives a basis for cooperation across nations .

He urged governments, academia, and companies worldwide to form an international community or consortium of AI safety research institutions dedicated to this challenge . Such a global network could collaboratively develop and share the technical means to make AI “benevolent and controllable” . Hinton suggested that while nations might be reluctant to share technologies that make AI more intelligent or competitive, they should be willing to share methods to make AI safer and more “friendly” – since all humanity would benefit . In an interview with China’s Yicai Media, he pointed out that countries may have diverging interests in the AI race, but “one core consensus can bring all nations together: preventing AI from replacing humans.”  Achieving that common goal, he argued, transcends geopolitical rivalries.

Hinton’s calls align with broader AI governance initiatives launched at WAIC. At the opening ceremony, officials announced the formation of a Global AI Governance Initiative Center and an “AI Governance Action Plan,” with United Nations Secretary-General António Guterres delivering remarks via video . Notably, Hinton himself participated in the International Dialogues on AI Safety held in Shanghai just before WAIC, where he and 20 other experts jointly signed the “Shanghai AI Safety Consensus” – a pledge to collaborate on mitigating AI risks . These moves underscore that policymakers and scientists are beginning to coordinate responses to the ethical and safety issues raised by advanced AI. Ensuring AI is developed “向善” (for good) and for the benefit of humanity was a recurring theme throughout the conference .

Other speakers reinforced the message. In the U.S.-China tech dialogue, Eric Schmidt praised open-source AI development (noting China’s leading models tend to open their weights, unlike most U.S. models) and argued for built-in “guardrails” to align AI with human values during training . Schmidt said he remains an AI optimist, but cautioned that as AI systems begin to self-improve in the next few years, it is critical to ensure “these processes do not spiral out of control” . “This is a problem faced by all humanity – every nation must come to the table to negotiate and cooperate,” Schmidt said, expressing hope for U.S.-China collaboration to prevent AI’s negative impacts while harnessing its benefits . Such sentiments echoed Hinton’s plea that global trust and coordination are essential to manage AI’s rise.

Human Minds, Machine Minds – Hinton’s Take on the Future of AI

Beyond risk warnings, Hinton also offered insight into the nature of intelligence itself and the promise of AI. In a thought-provoking aside, he suggested that the way human brains process language is remarkably similar to how large language models (LLMs) do. In fact, “human beings might themselves be large language models,” Hinton observed, drawing laughter and reflection from the audience . He noted that humans, like AI chatbots, can experience “hallucinations” – moments of confabulation or misunderstanding – though he acknowledged that AI systems have some fundamental differences and even strengths compared to humans . For example, Hinton pointed out that an AI model can share its knowledge instantly and replicate itself, and its “digital mind” isn’t limited by a mortal lifespan . A modern AI system running on servers can theoretically live and learn indefinitely, whereas a human brain cannot. These differences suggest that advanced AI might rapidly accumulate and disseminate knowledge in ways people never could.

Hinton struck an optimistic note about AI’s potential to advance science and society, provided it is properly guided. He highlighted how AI has already delivered groundbreaking results in fields like biology. One striking example, he said, is DeepMind’s AlphaFold, which used AI to solve the 50-year grand challenge of predicting protein structures – a feat that “greatly improved prediction accuracy” and arrived as an “early sign” of AI’s power to drive scientific progress . This success in protein folding, led by Demis Hassabis and colleagues, foreshadows AI unlocking new breakthroughs in many scientific domains, Hinton noted . From drug discovery to materials science, AI systems could vastly accelerate research and innovation. In Hinton’s view, the future of AI is not only about managing existential risks, but also about seizing these opportunities to better the world.

Before concluding, Hinton shared a bit of personal advice for the next generation of researchers. Speaking to young AI scientists in the audience, he encouraged them to pursue original ideas and not be afraid to challenge established consensus. “If you want to do truly original research, look for areas where you think everyone else is wrong,” he said – even if it means swimming against the tide . Breakthroughs in AI, Hinton suggested, often come from outsiders who refuse to give up on a bold idea. It was a fitting message from a pioneer who himself stuck with neural network research in the 1980s when the field was out of fashion, ultimately contributing to the deep learning revolution.

Balancing AI’s Promise and Peril

WAIC 2025 captured the dual realities of today’s AI landscape: astonishing technological progress coupled with profound societal concerns. Geoffrey Hinton’s presence – from his keynote warnings about superintelligence to his hopeful vision of global cooperation – embodied the conference’s central narrative. On one hand, AI’s leading minds are urging vigilance about the existential risks and ethical dilemmas posed by advanced AI. On the other, they are rallying stakeholders worldwide to ensure AI is developed responsibly so that its tremendous potential can be realized safely.

As the WAIC theme “Intelligent Era, Together for One World” suggests, the international AI community is increasingly united by a common destiny . Hinton’s call to action at WAIC 2025 – to “prevent AI from dominating humanity” by working collectively on AI safety, regulation, and ethics – resonated with attendees and policymakers alike . The consensus emerging from Shanghai was that no country can tackle the challenges of AI alone. Global cooperation and shared values will be essential to ensure that AI evolves as a positive force, not a “tiger” that humanity cannot control.

In Hinton’s own words, the overarching goal is clear: to stop artificial intelligence from ever “replacing or subjugating humans,” and to keep it as a powerful tool for good  . His message, delivered in China for the first time, reinforced a sense of urgency and unity in the AI community. For a world standing at the threshold of the AI age, the discussions at WAIC 2025 – and Hinton’s insights – may well shape the next steps in safely navigating the future of artificial intelligence.