New Delhi: Artificial general intelligence (AGI) could be years away, but its profound impact in the coming decade will be defined by scientific breakthroughs, not job extinction, Google DeepMind chief Demis Hassabis said at the India AI Impact Summit. “We’re at a threshold moment where AGI… is on the horizon, maybe in the next five to ten years,” said Hassabis. Yet he cautioned: “No, I don’t think we’re there yet.” For India, Hassabis suggested applying AI deeply in areas of existing strength, from agriculture to creative industries, and urged young people to become fluent in AI tools, which could make them “almost superpowered.” He also set a high bar for AGI. A true system must exhibit all human cognitive capabilities, across creativity, long-term planning and consistent reasoning across tasks. Today’s models, he said, remain “very impressive” but flawed. These models cannot continually learn once deployed and struggle with long-term coherent plans and struggle with simple tasks.
‘Cautious optimism, governance is key’
The nearer-term shift, he argued, lies in science. “The reason I spent my whole life and career working on AI is I saw it… as maybe the ultimate tool for science,” he said. Much of research involves detecting patterns in vast datasets, a task well suited to AI systems. He pointed to AlphaFold, Google DeepMind’s protein structure prediction system, as evidence that AI can “massively speed up our research and accelerate scientific discovery across almost any subject area.” Over the next decade, he expects a “new golden era for scientific discovery.” Humans, he stressed, remain central. “The next phase is going to be incredible… for human experts and scientists,” particularly in cross-disciplinary work. Fully autonomous “co-scientists” remain more than a decade away. “In AI systems today, it’s the human using it, putting in the energy and the question. Even in science, we have to supply the hypothesis and what the right research question is and then these tools help us get to those answers more quickly,” the Nobel prize awardee explained. Still, AI is “dual purpose,” he warned. Near-term risks include cyber threats and the misuse of AI in sensitive biological research.
