Dr. Phil Yang's AI Series 48. Conversational AI vs Physical AI: Why Don’t Robots Do What We Want?
- 양필승
- Feb 9
- 1 min read
Why does conversational AI feel so intelligent, while robots still struggle to move the way we want them to?
In Dr. Phil Yang's AI Series, Episode 48, we explore a fundamental distinction that often gets overlooked:
“Automation of intelligence” vs. “automation of the physical world.”
Intelligent AI operates in a probabilistic domain.When humans intervene—saying “this is wrong, try thinking this way”—the system adjusts probabilities and evolves. This process is often referred to as tweaking.
Physical AI (robotics) lives in a deterministic world.A deviation of even 1 millimeter when grasping an object can lead to failure or damage. As a result, unexpected human intervention is treated as noise or risk, and flexibility is intentionally sacrificed for mechanical safety.
This also explains Moravec’s Paradox: Tasks that are effortless for humans—walking, grasping, sensing—are among the hardest problems for machines.
So why is Physical AI suddenly back in the spotlight?
Because foundation models are beginning to be integrated into robotics, reshaping what is possible at the boundary between cognition and action.
A key takeaway from the episode:
With intelligent AI, humans are co-authors, building reasoning together.With physical AI, humans are still closer to supervisors than collaborators.
🎥 Watch the full episode here:https://youtu.be/SGlthsD7Cf4
Comments