The CEO of Google DeepMind, Demis Hassabis, predicts that the advent of artificial general intelligence (AGI) could occur as soon as 2030, give or take a year. Speaking at an event hosted by the Stanford Graduate School of Business, Hassabis characterized AGI as a pivotal milestone when AI will be able to handle a wide array of intellectual tasks, possibly surpassing human capabilities.
Hassabis described this forthcoming era as groundbreaking, stating that in a decade, society may look back and realize it was on the brink of a monumental shift. He highlighted that 2026 would mark a significant turning point, with AI systems becoming increasingly effective in various jobs, offering developers clearer insights into what remains to be done before AGI is achieved. He stressed that the responsibility of preparing for AGI shouldn’t rest solely on technologists, as its implications for society are profound.
He emphasized the urgent need for societal awareness and preparation regarding the implications of AGI, acknowledging that the near future will be crucial in shaping how this technology will impact lives. His remarks come amid ongoing debates within the AI community regarding how imminent AGI really is.
OpenAI CEO Sam Altman stated that the company possesses the knowledge to create AGI in the traditional sense, suggesting that AI could begin integrating into the workforce soon. Other prominent figures like Elon Musk and Dario Amodei of Anthropic have also made predictions about AGI emerging within the next few years. Musk projected that AGI could hit by 2026 and asserted that by 2030, AI may surpass the collective intelligence of humanity.
Conversely, some contend that AGI may have already been achieved, with current advanced models meeting its requirements. Eliza Labs' founder Shaw Walters claimed that we are at a pivotal moment and believe today's systems exhibit general intelligence.
Yet, there are skeptics who point out that existing AI still falls short of human-level reasoning. A report from the ARC Prize Foundation noted that leading AI models scored poorly on their benchmark test designed to assess adaptability in new environments, while humans achieved perfect scores, indicating substantial gaps in current capabilities.
The discussion surrounding AGI is further complicated by differing definitions within the AI field. Malo Bourgon, CEO of the Machine Intelligence Research Institute, mentioned that inconsistent criteria make it challenging to agree on when AGI has been realized. He remarked on the difficulty of defining what exactly qualifies as AGI in this evolving conversation.
Despite these debates, Hassabis remains optimistic about the rapid acceleration of technological progress. He predicts that the next decade will bring changes that surpass many people's expectations.
