Physical AI

The term Physical AI was systematically articulated by NVIDIA CEO Jensen Huang in his keynote speech at CES 2024, referring to AI systems capable of understanding and acting upon the physical world.

Physical AI refers to the technological direction of taking AI beyond screens and into the real world to interact with the physical environment, with typical deployment scenarios including robots, automated equipment, and edge sensing systems.

Work that requires deep creativity and fundamental breakthroughs is harder to automate with AI, while standardized, highly repetitive application-layer engineering is more likely to be replaced by AI tools

Jensen Huang has repeatedly emphasized in recent years that robotics and Physical AI represent the next trillion-dollar market.

The key difference between physical AI and purely software-based AI is that its decisions are embodied not only in text or voice output but also in embodied actions such as displacement, approach, staying, and turning, and these actions themselves carry social meaning.

The evaluation dimensions for physical AI social capability can include: unwelcome approach, forgotten deferred interaction, whether correction is accepted, and the proportion of appropriate responses.

A robot that starts speaking when it detects a person may only be reacting to "the person's presence" while ignoring "the person's situation".

Social design for physical AI can adopt a design approach that splits person detection, determination of a candidate invitation, whether conditions for responding are present, and execution of the response into four stages.

The AI design ideas discussed in the text are all tentative hypotheses, have not yet been validated through productization, and do not mean that the system already possesses human-like emotional experience.

For practitioners hoping to enter the field of physical AI, understanding robot operating systems, sensor fusion, reinforcement learning, and edge deployment will become just as important as mastering deep learning frameworks.

Physical AI systems must complete inference on the robot body or on nearby edge devices, for reasons including unacceptable wireless communication latency, limited bandwidth, and privacy and security requirements.

Physical AI systems need to solve three core challenges at the same time: real-time performance, safety, and generalization capability.

The core challenge of physical AI is bridging the gap between digital reasoning and physical execution, including understanding three-dimensional space, real-time sensor fusion, and built-in compliance with physical constraints.

SiMa.ai focuses on edge computing processors (physical AI chips) for real physical world application scenarios.

Vantora has explicitly stated its intention to go all-in on the Physical AI direction and is deploying the entire $100 million in funding toward that direction.

Physical AI commercialization faces challenges including hardware costs, safety and reliability, and the complexity of industrial environments, which will slow down the pace of commercialization.

Industrial manufacturing, logistics and warehousing, and energy infrastructure are currently the three vertical sectors with the fastest physical AI penetration rates.

Physical AI typically relies on multimodal perception (sensor fusion including vision, tactile, LiDAR, etc.), real-time reasoning, and low-latency closed-loop control with mechanical actuators.

NVIDIA CEO Jensen Huang has referred to physical AI as the next wave of AI

The research team believes that true physical AI must be able to work under zero-shot conditions, i.e., managing long-horizon physical tasks without task-specific training.