Summary

  • Demis Hassabis, CEO of Google DeepMind, predicts that artificial general intelligence (AGI) could be realized by around 2030, give or take a year.
  • He cautioned that society has limited time to prepare for the significant economic and social impacts of AGI.
  • His insights come amid ongoing discussions within the AI sector regarding the timeline for AGI's development.

Demis Hassabis, the CEO of Google DeepMind, suggests that the advent of artificial general intelligence is imminent, rather than a far-off aspiration.

During a recent event at Stanford Graduate School of Business, he expressed his belief that AGI, which refers to AI systems capable of performing various intellectual tasks at or above human proficiency, is likely to emerge by the decade's end.

"We have been referring to AGI as the next iteration of truly general artificial intelligence," Hassabis stated. "I am convinced that we are only a few years away from that, perhaps around 2030, plus or minus a year, which is truly remarkable to consider."

He described this period as the onset of a "new human era."

"When we reflect on this era a decade from now, we may realize that we were just at the foothills of the singularity," he commented.

Hassabis identified 2026 as a pivotal year, where AI tools began to demonstrate real utility in various professions, providing developers with a clearer path toward achieving AGI. He emphasized that preparing for AGI's impact cannot solely rest on technologists.

"Society needs to be aware of this, as we do not have much time to get ready for its implications. It will be incredibly significant," he remarked. "The future remains unwritten, but the upcoming years will be crucial in determining its direction and how we collectively envision it."

His statements come at a time when the discourse around the timeline for AGI's arrival is intensifying.

In the previous year, Sam Altman, CEO of OpenAI, claimed that the organization understands how to construct AGI as it has traditionally been defined, hinting that AI agents might soon start entering the workforce. Predictions from other industry leaders, including Dario Amodei of Anthropic and Elon Musk of SpaceX, also suggest that AGI could be achieved within the next few years.

Musk stated, "I believe we will reach AGI by 2026," during a December interview with Peter Diamandis, executive chairman of the XPRIZE Foundation. "I’m confident that by 2030, AI will surpass the collective intelligence of all humans."

Conversely, some experts assert that AGI has already been attained, with existing advanced models meeting its criteria.

"I believe we are at a turning point where AGI exists," remarked Shaw Walters, founder of Eliza Labs, in a prior interview. "I firmly believe that this represents general intelligence."

However, skeptics highlight that current AI systems still fall short of human-like reasoning capabilities. In March, the ARC Prize Foundation unveiled its ARC-AGI-3 benchmark, which assesses whether AI can learn and adapt in new environments. Top models from Google, OpenAI, Anthropic, and xAI scored below 1%, while human participants achieved perfect results.

The absence of a unified definition of AGI further complicates the conversation. Malo Bourgon, CEO of the Machine Intelligence Research Institute, pointed out that divergent definitions make it challenging to ascertain when AGI has actually been reached.

“There are numerous definitions to consider,” Bourgon stated. “When we discuss whether this system or that qualifies as AGI, it becomes quite complex to delineate what precisely meets the criteria for AGI.”

Despite this, Hassabis maintains that technological advancement is accelerating rapidly.

"The next decade will bring changes that likely exceed current expectations," he concluded.