Artificial intelligence pioneer Geoffrey Hinton has once again expressed deep concerns about the pace and direction of AI development, warning that many leading tech companies continue to downplay the potential risks involved. Speaking on the One Decision podcast, Hinton criticised the lack of genuine urgency among industry leaders, despite a growing understanding of the possible consequences.
“Many of the people in big companies, I think, are downplaying the risk publicly,” he said. While Hinton acknowledged that some figures within the AI community are aware of the dangers, he suggested this awareness is often not reflected in their public stance.
Hinton pointed to Demis Hassabis, CEO of Google DeepMind, as someone who recognises the seriousness of the issue. According to Hinton, Hassabis “really wants to do something about” the possibility of AI being exploited by malicious actors. DeepMind, founded in 2010 and acquired by Google four years later, remains central to the company’s AI research efforts.
The latest remarks build on earlier warnings issued by Hinton. In a separate interview on the Diary of a CEO podcast, he predicted large-scale displacement of white-collar workers due to AI systems. He described “mundane intellectual labour”, routine office and administrative work, as especially vulnerable, suggesting that AI could perform tasks previously handled by multiple people.
“I think for mundane intellectual labour, AI is just going to replace everybody,” he said, adding that automated systems could soon outperform humans in such roles with ease.
By contrast, he said blue-collar occupations remain relatively insulated from immediate threat, noting that roles involving physical labour would be more difficult to automate in the short term. However, Hinton cautioned that even this may not hold true indefinitely.
His comments come amid ongoing global debate about how to regulate artificial intelligence and mitigate its potential impact on jobs, privacy, and safety. Despite repeated calls for stronger safeguards, concrete policy measures remain limited.