Is AI repeating social media’s mistakes? Google DeepMind CEO Demis Hassabis warns against ‘move fast and break things’ mentality

As artificial intelligence reshapes industries and daily life, Google DeepMind CEO and Nobel Laureate Demis Hassabis is sounding a warning. Speaking at the Athens Innovation Summit alongside Greek Prime Minister Kyriakos Mitsotakis, Hassabis cautioned that AI must not replicate the same harmful patterns that have plagued social media platforms for years. “We’ve got a lot of complicated things to navigate,” he told the audience, emphasizing the need for responsibility over speed.

Learning from the Past: The Social Media Trap

Hassabis pointed out that Silicon Valley’s “move fast and break things” mindset led social media companies to prioritize user engagement without fully understanding its long-term consequences. “We should learn the lessons from social media, where this attitude went ahead of understanding what the consequent second- and third-order effects were going to be,” he said, as reported by the summit’s transcript.

He warned that AI models designed to maximize attention could amplify issues like addiction and mental health crises. “Social media algorithms grab more and more of your attention, but not necessarily in a way that’s beneficial to you as the individual,” Hassabis explained.

AI’s Promise and Its Perils

While AI offers opportunities to help users by summarizing information and filtering useful content, Hassabis cautioned that misuse could worsen existing problems. He called for international cooperation and regulatory standards, stating that governance must focus on tangible effects as AI becomes embedded in everyday products and services.

The stakes are high, he argued, because AI’s reach extends far beyond entertainment platforms, influencing healthcare, science, and global productivity. Striking a balance between bold innovation and risk mitigation, he believes, is a “continual tension” that will remain relevant as technologies move toward artificial general intelligence (AGI).

Early Signs of Trouble

Studies already suggest that AI may mirror social media’s toxic patterns. A University of Amsterdam experiment involving chatbots on a simulated network found that bots formed echo chambers and boosted extreme opinions without any external interference. Attempts to disrupt these patterns failed, suggesting that the underlying problem runs deeper than algorithms.

Hassabis urged technologists and regulators alike to adopt a scientific approach: test thoroughly, understand implications, and prioritize people over profits. “The rubber meets the road,” he said, as AI’s impact will ultimately be judged by how it is deployed in real-world services. His message is clear—AI’s promise is immense, but only responsible stewardship can prevent repeating social media’s mistakes.