On my vacation in Greece starting July 17th, I saw it as a good opportunity to see how artificial intelligence is perceived in this small European Union country at the intersection of Africa, the Middle East and Europe. My curiosity was amplified by the fact that I was far from famous islands like Santorini and Mykonos.
The first four days of my trip took me to Ionina, a town in the country's northwestern part of the country, a region of climbers and shepherds. Here, unlike the tanned islands to the south, the landscape is dominated by lush green forests and imposing mountains. The heat wave cleaning Greece in mid-July is preventing many tourists from tackling hiking trails in the area. Locals also told me that the ongoing war between Israel and Hamas has emptyed an area of thousands of Israeli tourists, usually visited every summer. In Ionina, if you are looking for fish, the waiter will give you an appearance saying, “What are you talking about?” This is a reminder that it is not Greece, the mythical beach, but the world in itself.
This remote feeling made me even more curious. Was the technology that dominates all the conversations in New York, America's leading companies, Wall Street even the topic of discussion here?
On Monday, July 21st, after admiring the sunset over the lake, where the residents of Ioanina gather, I found myself in a restaurant with a friend of a friend. There were four of us, including two locals. He is an engineer named Aristotelis Ifantis, a doctor named Thomas Zimus, a veteran doctor at a local public hospital. After holding back the first part of the nearly four-hour meal, I turned to Dr. Tzimas and asked if he had used AI.
He smiled. I immediately felt like I had asked the wrong question. Just because my work revolves around AI doesn't mean that everyone else will. But I was wrong. I just stumbled upon one of the most knowledgeable doctors I've ever met. Here, in a small town surrounded by Greek mountains, AI had already infiltrated a public hospital. Not only did he use it every day, he was keenly aware of its flaws and developed his own methods to counter them. His secret was telling me to control the “temperature” of AI and prevent him from inventing facts. Dr. Tzimas, 56, is an internal medicine specialist. He is “G. Hatzikosta,” director of the internal medicine department at General Hospital Ionina.
I might tell you soon too: Don't start Dr. Tzimas with AI. With his gentle, measured voice, he provides an endless stream of analysis on the subject. He introduces daily hospital routines and offers an intriguing glimpse into how large-scale language models are adopted here.
First, Dr. Tzimas dismisses the most common fears. “No, not at all,” says the 56-year-old, asking if I'm worried about him being replaced. “My job includes experiences that AI can't replicate yet. I need to listen to the lungs, heart, intestinal sounds, palpation patients and look into them.”
He continues, “AI may analyze digital photos of skin lesions, but it cannot palpate the patient and cannot sense pain or strength of security. These are quality that depends on the physician's senses, such as smelling the patient's breath to detect diabetic ketosidosis.”
The key he said was human interaction. Observing patient attitudes and hygiene provides important information that cannot be easily processed by the algorithm. “AI helps save time,” he admits. “However, you need an experienced doctor to enter these important details.”
He is also a pragmatist. Health professionals who don't need human touches claim they are already on the path to being replaced. The first on his list is radiology. He says that experts interpreting x-rays and MRI have become obsolete because AI has become so much better at finding pulmonary nodules, fractures and subtle neurological conditions. Studies published in journals like the Lancet show AI models that match or perform human radiologists when identifying mammogram cancer.
“In the case of X-rays, there is only one [specialist] To check 10 of them to replace them and to see what the system diagnosed,” Dr. Tzimas said. “And it's easy to do with CT and MRI scans. These specialties will soon be replaced by AI.”
He then says he is a psychotherapist. The explosion of mental health apps and chatbots like Woebot and Wysa already offers AI-driven therapy services. Dr. Tzimas believes these tools are poised to compete directly with human therapists for basic counseling, especially for patients who cannot afford or easily access in-person care.
“In psychoanalysis, you don't prescribe medication,” he explained. “Psychiatrists with severely schizophrenia or depressed patients need to prescribe medication. In treatment, you talk to them. AI will replace therapists.”
Before I could handle it, Dr. Tzimas revealed that some surgeons were also in the chopping block. He refers to human operators of surgical robots, such as the Intuitive's Da Vinci system, a multi-armed machine that allows surgeons to perform minimal invasive procedures more accurately. He believes it is only a matter of time before AI directly controls these robots, and it is only a matter of time before they allow the operation to be performed autonomously.
“Did you know there are robots that have been undergoing surgery recently? But they are controlled by humans,” he said, looking at my surprise appearance. “Yeah! They can be controlled by an AI system. That's coming.”
For now, AI has become his invaluable assistant. “My previous boss will write his thoughts on a piece of paper and his secretary will enter them,” he says. “Now I will tell my iPhone the notes, copy them and paste them into an AI system, generate professional emails, and also transfer registered documents from the referral to a digital archive, saving time.”
He says one of the most complicated parts of his job is navigating politics among medical staff that requires a delicate balance with diplomacy. This is where AI shines. “I use the 'negotiator GPT' prompt to make the answer very diplomatic,” he says. “If there is conflict between junior doctors, I need to be strict but prevent further conflict. This AI helps me to behave like a skilled diplomat and create replies on a communications platform that smooths out the issues.”
He also uses it to coordinate patient care after a hospital visit. “I think certain aspects of the role of a nutritionist, especially the creation of a detailed nutrition plan for patients, can be largely automated,” Dr. Tzimas said. “Imagine a patient with acute cholecystitis, who should not eat fatty foods, but who don't know what “fatty foods” mean. By encouraging nutritional AI, you can issue detailed meals.
However, the biggest impact that AI has on his job is the diagnosis of the patient's symptoms. Dr. Tzimas uses a recent example of a patient who has come to the hospital with a fever after inhaling dust into a stable sheep and goat goat. After entering symptoms and results from the basic test, the AI identified Q Fever as a potential diagnosis. A blood sample was sent to Athens for a specialized test. “It went back negatively and the final diagnosis was irrelevant, but the AI system identified Q Fever as a potential possibility, which opened our sights. Without AI, Q Fever might not have been on our list.”
His biggest concern is the composition of AI. He soon realized that AI hallucinates or makes things too often. However, this tool is too important to dispose of. Therefore, to limit these production, he focuses on a specific setting called “temperature.”
“If the temperature of the AI is 1, they'll hallucinate,” he explained. “In the medical field, it must be 0.3. It makes them very strict. They can't fantasize about things. They stick to the facts.”
Temperature is a parameter in the AI language model that controls randomness. A high temperature of nearly 1.0 encourages AI to become more creative and unpredictable. This is great for writing poetry, but is potentially dangerous to diagnose illness. Bringing the temperature closer to 0.3 reduces the risk of these hallucinations as AI is more concentrated, deterministic and factual.
“When you encourage AI with very strict protocols, they don't hallucinate,” Dr. Tzimas said, as I forced the reliability of his tools onto him.
In Ioanina, for Dr. Tzimas, the debate about whether AI changes drugs is irrelevant. I already have it. He uses it to diagnose, communicate with staff, and even teach junior doctors. His perspective reveals both the promises and dangers of AI's future. On the one hand, treatment, radiology, and surgery do not look the same. On the other hand, the difference in temperature between 1.0 and 0.3, which is a simple software configuration, can be a difference between correct diagnosis and catastrophic errors.
I was left with dinner. Seeing small town doctors adopting AI so aggressively, I thought the gap between technology and what I don't have would grow at a slower pace than I was afraid. The gap didn't seem to be a bit big, at least in Ioannina.
