DeepMind CEO: AGI May Arrive in 5–10 Years But AI Still Lacks Imagination

Google DeepMind CEO Demis Hassabis is of the opinion that artificial general intelligence (AGI)—AI systems with human-like cognitive abilities—may be a reality in the next five to ten years. But he warns that present AI is not imaginative, creative, or conscious in any way.

Speaking during an interview on CBS News' 60 Minutes on April 22, Hassabis explained that although AI has made quick progress, it still can't pose genuinely new questions or create novel hypotheses outside of what it's learned.

AI Still Doesn't Think Like Humans, Says DeepMind CEO

"They're still the average of all the human knowledge out there," Hassabis said, pointing to the limitations of current AI models. "They can't yet go beyond and create something completely new."

Hassabis, who holds a Ph.D. in neuroscience and co-created AlphaFold, an AI program that can predict protein structures, stressed the importance of responsible development:

"My advice would be to create smart tools initially and use them to assist us in driving neuroscience forward before we reach the point of contemplating things like self-awareness."

AI May Revolutionize Drug Development

As Hassabis sees it, AI may radically shorten drug development timelines, potentially releasing new treatments within months rather than years.

"Creating a drug today costs about 10 years and billions of dollars. AI might reduce that to weeks," he asserted.

Humanoid Robots and the Future of Robotics
The CEO of DeepMind also indicated robotics as the future. "We might see within a few years of advances in humanoid robots or similar types that know and can interact with the physical world," he forecast.

A Knighted AI Pioneer

Hassabis, who attended Cambridge, MIT, and Harvard, was knighted by King Charles and received the 2024 Nobel Prize in Chemistry for his research on AlphaFold, together with fellow scientist John Jumper.

Also Read: iQOO Z10 Turbo Pro: 7,000mAh + 120W Charging Launches April 28

AI Still a 'Black Box'

Although its capabilities are increasing, Hassabis admitted that AI is still unpredictable. "We don't program these abilities directly; they emerge from the data," he said, demanding guardrails and safety measures as the technology evolves.

“We are on an exponential curve of improvement, and that calls for thoughtful development,” Hassabis concluded.