DeepMind CEO predicts human-level AI within a decade

CEO Demis Hassabis (photo) of Google DeepMind, Google’s artificial intelligence (AI) organization, predicted that AI that could match humans at any task will be here within the next five to 10 years. Sam Altman, CEO of OpenAI, the developer of ChatGPT, also predicted that “it could be developed in the relatively near future,” referring to the time frame as specified by Hassabis of within five years at the earliest.

Speaking at a media briefing held at the Google DeepMind headquarters in London on Monday (local time), Hassabis said that AI functions will start coming to the fore in the next five to ten years and advancing to ‘artificial general intelligence (AGI). He won the Nobel Prize in Chemistry last year in acknowledgment of his contribution to developing the AI ‘AlphaFold’ predicting protein structures.

“AGI is a system that can perform all complex tasks that humans can. Current AI systems are still passive and cannot do many things, so they have not reached the AGI stage,” Hassabis said. “A considerable amount of research is needed to reach the goal (AGI).” He defined AGI as a system that can exhibit all the complicated capabilities that humans can. “I think today’s systems are very passive, but there’s still a lot of things they can’t do. We’ve still got quite a lot of research work to do,” Hassabis said.

Hassabis said that the main challenge with achieving artificial general intelligence is getting today’s AI systems to understand the context from the real world. While it’s been possible to develop systems that can break down problems and complete tasks autonomously in the realm of games, such as the game Go, bringing such technology into the real world is proving harder.

Regarding artificial superintelligence (ASI), AI with intellectual abilities that surpass those of humans, Hassabis said, “ASI comes after AGI surpassing human intelligence. No one knows when such disruption will happen.”

Hassabis’s prediction for the rise of AGI is somewhat later than the predictions of other big tech giants. Last year, Tesla CEO Elon Musk said that AGI technology could be developed “by 2026,” while Dario Amodei, CEO of Anthropic, a U.S.-based AI company known to rival “OpenAI” predicted that “within the next two to three years, AI could outsmart humans in most tasks.”


장은지기자 jej@donga.com