News Update

Sometimes the reporting on artificial intelligence makes it sound as though its chief threat is to the jobs of assembly line workers, law library clerks, and, someday, maybe truck drivers. Those certainly are trends. You could say that AI’s “entry level job” was to displace entry-level workers. That probably has some truth if we are dealing with sheer numbers, but it should not imply that AI is merely advancing up the workplace ladder from the bottom rung.

The BBC reported on August 28, that surgeons at University College’s London Hospital performed the first brain surgery with live AI and, in part, attributed its success to AI’s role.  In brief, surgeons were removing an 11 cm pituitary tumor from 48-year-old Rhys Hibbert, a tumor that threatened his vision. AI analyzed a live video feed of the operation in real time instead of relying only on pre-surgical scans. That enabled surgeons to avoid hidden, critical nerves and blood vessels for safe removal of the tumor. The patient’s sight was saved. The tool analyzes live endoscopic feeds to map out anatomy in real time; but it also uses facts, analysis, and projection to predict the surgeon’s next movements 15 to 30 seconds in advance. And alerts the surgeon to prevent accidental cuts.

Lest you conclude that this is just computer vision and projections read on the screen by surgeons, the chief of surgery at a clinic in Santiago, Chile, recently reported that an AI guided camara enabled him to perform gallbladder removal surgery alone. No colleagues, no residents, no nurses. The procedures combined magnetic surgical instruments with software that guided the surgical camera — usually the job of an assistant. The camera traced the surgeon’s tools inside the body of the patient and kept adjusting to get the best angles. Dr. Ricardo Fune said: “The camera was following me wherever I moved my hands and the whole process was excellent. This camera lets us do the surgery alone.”

Robots in the OR

Arguably, this is the cutting edge of AI’s role in the operating theater, but already the global surgical robot market is estimated to reach $64.4 billion by 2034 according to statistics from Precedence Research.  An expert on surgical robotic technology, Bhusham Jayeshkuman Patel, wrote in Forbes that robotic surgery “is not just a trend but a fundamental shift that aims to elevate minimally invasive procedures to new heights.”

For now, AI is most common in complex, image-heavy, or minimally invasive procedures: for example, orthopedic joint replacements (navigating bone density), urology and gynecology (robotic prostatectomies/hysterectomies), and oncological tumor resections where millimetric precision is vital.

More broadly, the applications of AI in the operating room today are computer vision and real-time guidance (the brain operation), stepwise task automation (automating repetitive, fatigue-inducing auxiliary micro-tasks like positioning surgical cameras or guiding robotic arms to suture tissue and clip blood vessels, and preoperative and postoperative analytics (reviewing vast amounts of information from patient charts, scans, and genomic data beforehand to build custom surgical roadmaps and post-surgery tracking recovery metrics to flag potential infections early).

The latest powerful trend in AI is toward humanoid robots but you will not find them standing around the operating table. Far more useful are fixed, stable-based robots with multiple spidery arms and “hands” able to achieve movements impossible to human wrists. These can clip blood vessels, suture, pass instruments, keep track gauze.… In short, the high-stakes theater of surgery belongs firmly to non-humanoid, highly specialized machinery.

Lowering Costs and Relieving Staffing

Operating rooms are a hospital’s prime real estate and AI can optimize every minute spent in them. With more than 51 million procedures in the U.S. each year, surgery is a major driver of costs, accounting for an estimated 34 percent of employers’ total healthcare spending. A report in the Journal of Robotic Surgery in 2025 said that AI-driven surgeries reduced operating time by some 25 percent and improved surgical precision by 40 percent, reflected in enhanced targeting accuracy during tumor resections and implant placements. Shorter surgeries mean more surgeries each day.

The same article reported a 30 percent reduction in intraoperative complications and reduced patient recovery times. Thus AI lowers the heavy costs of prolonged hospital stays for recovery and readmissions for postoperative problems.

There is a mounting shortage of medical staff in the United States with the National Center for Health Workforce Analysis projecting a shortfall of more than 141,000 physicians and more than 100,000 nurses by 2038. AI relieves human assistants and nurses from jobs like holding cameras and tracking gauze. Outside the OR, AI handles more conventional tasks with chatbots handling routine post-op patient queries at 1:00 AM, heavily filtering out non-emergency calls, and reducing the strain on on-call nursing staff. Here as elsewhere, it goes without saying, efficiency is achieved but not necessarily patient satisfaction when AI decides that no nurse is needed for the patient’s complaint of insomnia.

The ‘Insurance Wars’

Currently, insurance reimbursement, malpractice, and other items concerning payment and expense dominate our health care debates. As a  new technology in the surgical arena — a technology already a lightning rod for fears and complaints — the rapid deployment of AI in surgeries comes under intense scrutiny from regulators and the public.

Then there is the debate over malfunction and malpractice, which became a flashpoint after FDA adverse-event reports linked AI-assisted medical devices to botched procedures and the misidentification of body parts. Every patient’s anatomy is unique, and when an AI algorithm misinterprets what it sees, a single error can cascade into serious consequences. That raises a difficult question: When AI contributes to a medical error, where should legal liability fall?

There also is the well-worn charge of algorithmic bias. Are medical data sets representative of all possible demographics:  race, ethnicity, gender, and age? The question answers itself, thus critics charge that using an AI trained on a “narrow” data set can lead to dangerous “hallucinations” or errors when operating on a patient with atypical anatomy.

And then, if the younger generation, especially those still in school, are at grave risk of never learning to write or think critically, what about human surgeons? Will they eventually experience “skill decay”?  Or perhaps they will hesitate to override an automated system when it makes an incorrect high-stakes recommendation.

An article in Medicine looked at the AI insurance landscape and legal challenges associated with medical robots: their legal status, liability in cases of malpractice, and concerns over patient data privacy and security. What emerges from this and other commentary is a deeply fractured landscape sometimes framed as a battle between clinical denial and administrative weaponization.

In terms of clinical coverage (or “non-coverage”), commercial insurers lag AI technology. Long-term clinical data upon which the industry depends is still being gathered, so insurers routinely label advanced AI-driven robotic surgical techniques as “experimental.” That leaves up to 90 percent of licensed AI medical devices entirely uncovered by standard insurance plans. For example, patients are often caught in legal disputes over coverage of steep hospital upcharges for robotic assistance.

Ironically,  perhaps, insurance companies are aggressively adopting AI to manage their own costs. A highly controversial program initiated by Medicare uses AI models (like the WISeR model) to review and reject overused or low-value medical claims, giving the AI tech vendors a cut of the financial savings.

To end on a positive note, AI is streamlining the often-slow, frustrating, and even dangerous administrative “prior-authorization” process, reducing manual entry, and cutting the average claim approval cycle significantly.

The Residency Training of ChatGPT?

Questions of competence and reliability often come back to education and training. At this time, that field is ad hoc and experimental — as is the use of AI. In Science Robotics, researchers at Johns Hopkins describe how they trained their surgical robot (more formally “hierarchical surgical robot transformer),” SRT-H, to perform part of a gallbladder removal surgery in a lab. They created an AI model with natural language processing similar to ChatGPT for their robot. The robot was trained by watching videos of gallbladder surgeries and then operated on eight pig gallbladders. SRT-H responded to and learned from voice commands just as would a novice surgeon working with a mentor.

The “student,” or should we say “resident,” performed a string of 17 tasks — identifying ducts and arteries, placing clips and cutting with scissors. It adapted to individual anatomical features in real-time, made decisions on the fly, and self-corrected when things didn’t go as expected.  The report from the Hopkins Whiting School of Engineering said the robot performed the surgery with 100 percent accuracy.

The school put this spin on the report, quoting Axel Krieger, a “medical roboticist”:

“This … moves us from robots that can execute specific surgical tasks to robots that truly understand surgical procedures…. closer to clinically viable autonomous surgical systems that can work in the messy, unpredictable reality of actual patient care.

Prof. Krieger naturally adopts the “AI speak” universal in the field. The robot does not “truly understand surgical procedures” in any sense we mean; it truly “understands” nothing. An article in Frontiers in Public Health puts it this way: “AI does not involve sentience or consciousness but focuses on data processing, pattern recognition, and prediction through algorithms and learned experiences.”

A multi-step procedure like a gallbladder removal is a massive leap in engineering, which suggests that AI’s role in the operating theater is only beginning. But the AI surgeon is does not possess consciousness. Rather, the point we should take away is how much of what we view as human judgment and discrimination reduces to pattern recognition. We recognize a known pattern in a situation, including sometimes an emerging mistake, and act accordingly.

As the AI revolution unfolds in surgery and other fields, we may see with increasing clarity, in the mirror of artificial intelligence, what remains uniquely human.

READ MORE from Walter Donway:

Socializing and Longevity: Does Online Count?

Would the Whiff of AI Have Panicked Harold Ross?

READ MORE:

Why Aren’t Our Modern-Day Robber Barons Giving Us Robber-Baron Benefits?

Walter Donway was a health program officer for The Commonwealth Fund, program director for the Dana Foundation, and founding editor with the late William Safire of Cerebrum: The Dana Forum on Brain Science. He is a freelance writer, now, living in East Hampton, New York.  In March of this year, he published A Serious Chat with Artificial Intelligence (Romantic Revolution Books, 2026).