NVIDIA Director Of AI Jim Fan Proposes A ‘Physical Turing Test’

The Turing Test — the ability of a computing system to pass off as a human in a conversation — was once thought to be one of the biggest milestones for computing. This milestone has been conclusively passed by modern LLMs. But Jim Fan, the Director of AI at NVIDIA, has proposed a new test for computers to conquer — the physical Turing test.

“The Physical Turing Test: your house is a complete mess after a Sunday hackathon. On Monday night, you come home to an immaculate living room and a candlelight dinner. And you couldn’t tell whether a human or a machine had been there. Deceptively simple, insanely hard,” Fan wrote on X.

“It is the next North Star of AI. The dream that keeps me awake 12 am at the lab. The vision for the next computing platform that automates chunks of atoms instead of chunks of bits,” he added.

Like Fan says, the test, like the original Turing Test, seems quite simple, but could be quite complex to achieve. It would need for several different aspects to come together for a robot to clean a house and cook dinner. The robot would need to navigate a domestic environment, and have the ability to grasp at bedsheets and turn on knobs. It would need to be able to see and take feedback from the environment. It would also need to have the intelligence to understand commands and take decisions if things don’t go as planned — if say, a crucial ingredient is missing while preparing dinner.

And this could end up being the new problem for researchers to aim at in the coming years. The Turing Test was passed with almost no fanfare — LLMs were introduced to the world in late 2022, and quickly improved in capabilities until it was impossible to tell if you were talking to a human or an LLM. This process occurred so rapidly that no one could conclusively determine when the exact moment was that the test had been passed. It’s likely that the Physical Turing Test would go the same way, but there might need to be lots of technical breakthroughs in computer vision, robotics and general intelligence before it becomes a reality.

Posted in AI