Embodied AI Glossary中文Physical Turing Test
物理图灵测试CommonA robot passes if people cannot tell whether a piece of real-world physical work was done by a human or a machine.
The physical Turing test is a phrase coined by NVIDIA's senior director of AI, Jim Fan, in a talk at Sequoia Capital's AI Ascent conference in May 2025, moving the Turing test from “can't tell human from machine in conversation” into the physical world: have a robot do real physical work, such as housework, and if a person looking at the result cannot tell whether a human or a robot did it, the test is passed. In the talk, he argued for using large-scale simulation to make up for the shortage of robot training data, and described ideas such as digital twins and “digital cousins.” Similar ideas existed earlier: cognitive scientist Stevan Harnad's “total Turing test” likewise required testing both perception and the ability to manipulate objects. The phrase is now commonly used to describe the ultimate goal of a general-purpose household robot.
ExampleJim Fan's example scenario: your house is a mess after a party on Sunday night; you come home Monday to find it fully cleaned up, with a candlelit dinner set on the table, and you cannot tell whether a person or a robot did it.
- Related
- Embodied AI · Physical AI · General-purpose Robot · Household Tasks · Moravec's Paradox · Digital Cousin
- Sources
- The Physical Turing Test: Jim Fan on Nvidia's Roadmap for Embodied AI (Sequoia Capital, YouTube)
AI Ascent 2025 (Sequoia Capital)
Turing test - Wikipedia - As of
- 2025-05