Select Language:
Medical robots are poised to become one of the earliest commercial applications of embodied intelligence, though there are still hurdles to overcome before moving from laboratory demonstrations to widespread real-world use, according to a leading expert from Tongji University, who specializes in developing robots for heart procedures and blood collection.
Interest in “world models” and “physical artificial intelligence” is mainly driven by computer scientists seeking to expand AI into tangible systems. Meanwhile, the expert’s own work stems from robotics and control theory, aiming to integrate AI into machines he’s spent years developing.
Despite differing approaches, both groups share a common goal: fostering embodied intelligence—AI that can perceive, learn from, and interact with the physical environment in meaningful ways.
Unlike humanoid robots, which are still seeking scalable commercial applications, medical robots operate in markets with clearer demand, targeted users, and highly regulated, structured environments. AI is transforming these medical devices from highly precise tools into intelligent systems capable of better environmental perception and clinical support.
Different industries may follow various paths toward commercial success. Industrial robots often benefit from straightforward workflows, while medical robots face obstacles such as rigorous safety standards, regulatory approvals, and acceptance from healthcare providers.
The expert’s team has created a cardiac intervention robot that utilizes multiple imaging modalities in real time to assist surgeons during procedures. They also developed a blood-drawing robot that employs computer vision to locate veins and automatically adjust needle placement, boasting a success rate of over 98% on the first attempt.
While medical AI has made significant progress, it primarily aims to support healthcare professionals rather than replace them. The expert emphasized, “The robot can suggest actions, but the final decision still rests with the doctor.”
The biggest challenge for embodied AI is not just showcasing impressive lab results but ensuring consistent, reliable performance in complex, real-world settings. Although his cardiac robot performed well in simulations, early clinical trials revealed unexpected issues, including equipment interference and mismatches between robotic control systems and doctors’ practices.
The team worked closely with hospitals to improve the system, noting that “lab success doesn’t necessarily translate to clinical success.”
A broader discussion at this year’s AI conference centered on how quickly AI can adapt to physical environments. Industry leaders highlighted the divergence between rapid technological advancements and slower organizational changes—many industries see technology evolving quarterly, but companies take years to adapt.
Experts stressed that future AI must be capable of understanding and predicting changes in the real world. They also noted that AI is shifting from simply offering capabilities to becoming a true partner that delivers tangible results.
While human and machine intelligence differ fundamentally, machines excel at computation, whereas humans have advantages in emotional understanding and social interaction.
China’s strength in medical robotics lies in its rich clinical scenarios and a robust engineering ecosystem capable of swiftly transforming medical needs into market-ready products. However, for Chinese firms venturing abroad, building trust among medical professionals remains the biggest obstacle. Trust, they say, takes time to develop—surgeons accustomed to international platforms need concrete evidence that new systems can safely enhance clinical outcomes.
As embodied intelligence moves toward commercialization, success is increasingly dependent on systems that provide measurable benefits in real-world environments, rather than merely resembling human-like appearances.
ChatGPT
Perplexity
Gemini AI
Grok AI
