Demis Hassabis says AI could gain self-awareness within a decade

DeepMind CEO Demis Hassabis thinks machines might begin waking up sooner than many expect.

While today’s AI lacks true consciousness, Hassabis predicts that self-awareness could emerge "implicitly" within the next five to ten years.

From self-identity to scientific imagination

According to Hassabis, AI models will soon need to understand the distinction between "you, self, and other." He points to that basic boundary as the early foundation of awareness.

Right now, AI models still lack genuine imagination. But within five to ten years, Hassabis expects them to advance significantly: they won't just solve complex scientific conjectures, but pose brand-new ones on their own.

Why it matters

Coming from the leader behind AlphaFold, this timeline carries serious weight in the AI research world.

We think Hassabis takes a smart approach by skipping the vague philosophical debates about machine souls. Instead, he treats self-awareness as a practical technical milestone, moving from subject-object distinctions to active scientific curiosity.

If AI truly evolves from passive pattern matching to generating original hypotheses, researchers will stop treating it as a coding copilot and start working with it as an autonomous lab partner. Still, leaping from statistical prediction to genuine selfhood remains a massive, unresolved hurdle.