Le PDG de Google DeepMind a un message pour ceux qui craignent les suppressions d'emplois liées à l'IA

Google DeepMind CEO Demis Hassabis has delivered a message to those concerned about the potential impact of artificial intelligence on employment, emphasizing that while the technology poses challenges, it also presents opportunities for human ingenuity. Speaking at the Stanford Graduate School of Business, Hassabis outlined his vision for the future of artificial general intelligence (AGI), predicting it could arrive by 2030, give or take a year. He described AGI as an “enormous transformative technology” set to usher in “a new human era.” According to Hassabis, society will likely look back in a decade and realize it was “standing in the foothills of the singularity”, the hypothetical point at which AI surpasses human intelligence and begins improving itself autonomously. Hassabis acknowledged growing concerns about job displacement due to automation, noting that AI is already reshaping industries such as coding and daily task management. However, he stressed that AGI, which possesses the capacity for reasoning, decision-making, and adaptation, might not replace all roles. Instead, he highlighted the enduring value of human traits such as taste, design sensibility, original thinking, and interdisciplinary synthesis. During an interview with The Rundown AI’s Rowan Cheung, he urged individuals to recognize their unique abilities, stating, “We’re general intelligences ourselves, don’t forget. Look at what we built around us, it’s incredible, with our hunter-gatherer brains. Why would we stop here?” He emphasized that humans are “instantly adaptable” and that AGI will enable “amazing new things” rather than stifle human creativity. The AI community remains divided on the timeline for achieving AGI, with varying predictions from leading figures in the field. While Hassabis forecasts AGI by 2030, others offer different assessments. For instance, Sam Altman has suggested humanity is close to creating digital superintelligence, Dario Amodei predicts powerful AI could emerge as early as late 2026, and Jack Clark and Marina Favaro estimate that by 2027, AI systems might complete tasks that typically take humans weeks. Conversely, Yann LeCun, a prominent figure in AI research, has dismissed the notion of general intelligence, calling it “complete BS” and asserting that current transformer models lack the cognitive depth of human intelligence. Meanwhile, DeepMind’s chief AGI scientist, Shane Legg, estimates there is a 50% chance of “minimal AGI” emerging by 2028. In addition to these projections, Hassabis proposed a novel approach to regulating AI development. He advocated for a global watchdog led by the United States, suggesting the creation of a new standards body tasked with overseeing the testing of the most advanced AI models before they are deployed. This body, modeled after the Financial Industry Regulatory Authority, would operate as a public-private entity, funded primarily by industry stakeholders and staffed with independent technical experts and open-source advocates. Hassabis argued that traditional government agencies would struggle to respond swiftly or effectively to the rapid pace of innovation, making a self-regulatory framework essential. Under this proposal, AI models would be classified as “frontier-class” once they meet specific performance benchmarks, requiring developers to register as “frontier labs” and adhere to additional oversight protocols. This system aims to address regulatory debates by focusing solely on the capabilities of AI systems, regardless of their origin or openness. As the debate over AI regulation intensifies, Hassabis’s proposals highlight both the urgency of preparing for AGI and the need for coordinated governance. His emphasis on human adaptability and creative potential offers a counterpoint to fears of technological unemployment, while his call for a unified regulatory framework underscores the complexity of managing AI’s societal impact. The coming months will likely see further discussions on how best to balance innovation with responsibility, shaping the trajectory of AI development in the years ahead.