The Shift to Physical Intelligence: Why Healthcare is Abandoning the "Linguistic Illusion"
On December 18, 2025, OpenAI’s Sam Altman admitted that the industry is still in a "primitive era" regarding digital memory. While the tech press treated this as a roadmap update, for those of us working at the intersection of medicine and code, it was something else entirely: a tacit admission that the honeymoon between healthcare and Large Language Models (LLMs) is over.
As a medical doctor and a full-stack developer watching from both the operating room and the Linux terminal, I view healthcare as the primary victim of what I call the "Linguistic Illusion." For the past three years, the industry has poured billions into systems that excel at talking about illness but remain utterly incapable of understanding the patient.
At Medevel, we have been tracking this divergence closely. Here is our analysis of why the era of the "Verbose Physician" is ending, and why we are pivoting toward "Physical Intelligence."
1. The "Digital Placebo": Why Hallucination is a Fatal Flaw
The core issue with LLMs in healthcare is not that they make mistakes; it is that they are probabilistic mimicry engines designed to prioritize eloquence over veracity.
In the context of software development, an AI hallucination is a syntax error, annoying, but fixable. In medicine, hallucination is not a bug; it is a structural failure with life-or-death consequences.
When a patient asks an LLM about symptoms, they are not receiving medical insight based on causal reasoning; they are receiving a word-frequency score derived from scraped clinical texts.
Hazem Abbas
This creates a "Digital Placebo" effect, particularly in mental health. Current AI therapy apps pose an existential risk because they do not understand depression as a physiological state.
They merely predict the next token most likely to sound empathetic. They can mimic the rhetoric of care while remaining blind to vocal micro-tremors or behavioral shifts that indicate suicidal ideation. Doctors do not need rhetoric. We need truth.
Hazem Abbas
2. The Crisis of the Physical Gap
The narrative is shifting from a "Chat Economy" to a "Movement Economy."
The future of surgery does not lie in a robot that can write a poem about a scalpel or summarize an EHR report in Shakespearean English. It lies in a machine that understands physics: tissue resistance, micro-angle cutting dynamics, and fluid pressure.
This is the promise of World Models. We are transitioning from software that lives on screens to "Smartware" that interacts with biological reality. Betting on teaching AI "grammar rules" instead of "Newton’s laws" is investing in the past.
Hazem Abbas
3. $15 Trillion Looking for a Body
While Silicon Valley remains obsessed with digitizing information (automating reports, text summarization), it ignores the "elephant in the room": the $15 trillion physical economy.
Hospitals, drug logistics, and elderly care are not text-based problems. They are spatiotemporal challenges. Robots that navigate hospital corridors or assist seniors with mobility require intelligence that understands 3D space, time, and motion, not just a high-level advisor that repeats safety protocols. We need intelligence that acts with hands.
4. Open Source vs. Tech Feudalism
As an open-source advocate who has championed Linux since the 90s, this transition terrifies me.
Training on text was expensive, but training on physical simulation and world modeling, as NVIDIA and Tesla are currently doing, requires astronomical computational power.
The risk is clear: If LLMs created monopolies like OpenAI, World Models threaten to create a closed "Tech Feudalism." The existential question for us at Medevel and the wider developer community is whether we will see an "open-source Linux" for medical world models, or if doctors and developers will become mere tenants of the few tech giants capable of simulating reality.
Hazem Abbas
Conclusion: The Era of the Physical Surgeon
The age of data accumulation has reached diminishing returns. The solution is no longer "more text"; it is deeper understanding.
At Medevel, we are shifting our focus. We are no longer impressed by an AI's ability to write. The future belongs to those who build the most accurate physical model, one that truly understands the human body and the laws that govern it.
The revolution isn't in language. It is in motion.
Hazem Abbas
Hazem Abbas
