Dr. Rhonda Patrick: And I’ve heard you say something sort of interesting, which perhaps I’m not quoting directly, but that it kind of should be medical malpractice, in a way, for a physician today, right now, not to be using AI. Can you talk a little bit about why you said that, what it means for a physician to use AI responsibly, and also how patients can advocate for themselves? That’s another area.

Dr. Derya Unutmaz: Yeah. In fact, I said that after the o1 model came out. I think that was sort of the first reasoning model. I was testing a lot of medical questions. I have a medical degree, but I don’t see patients. I have a lot of friends, and I have some knowledge of how medicine works. Some of the questions are hard, and some involve real-time data. Before o1, it was great at reaching the literature. A physician might lack certain knowledge, and it knew what was published recently, but it was not at the reasoning level.

The o1 model was able to reason, and reasoning is extremely important in medicine. Even if you have all the information, you still have to consider that person’s context, what would be more likely to treat that person, and how to diagnose it. We don’t always know the answer. I think o1 was able to get to that point. At that point, I said it is now unethical for physicians not to use AI anymore. I didn’t say malpractice yet, but it is truly unethical in the sense that you can use it and still apply your judgment. It can prevent you from missing an obvious mistake, and sometimes a nonobvious mistake, or help diagnose conditions that require multiple clinical specialties to come together when you don’t have that capability. You might live in a village or something.

Now I feel that it is truly going to be considered malpractice, in my opinion. It is not legally so, but eventually it will be, because the current advanced models are able to diagnose and write a treatment protocol better than or as well as a specialist in that field. It’s not just a family physician. Let’s say you have a very complex cancer, you know the mutations and what’s not there, and you go to an oncologist who is very specialized in that. I believe the current models are at that level. Of course, not every specialist is a top specialist. If that were the case, we wouldn’t have millions of misdiagnoses and mistreatments in the United States every year.

I think they said something like 12 million misdiagnoses. I think 700,000 people suffer from them or die from them. Some are totally innocent. Any doctor could have missed them. But now AI wouldn’t miss that. Even a specialist might make a mistake, misdiagnose, or mistreat because they lack certain things that the model doesn’t. Imagine that you refuse to use an MRI or CT machine because you say, “That’s too much technology. I’m just going to do an X-ray because that’s enough for me,” and you miss a tumor. AI models are able to detect certain tumors, like breast cancer, years before a radiologist is able to see them. If you miss that, that person is going to die if you don’t know.

To me, that becomes malpractice because the technology is at that level now. Missing breast cancer wouldn’t have been malpractice five years ago because nobody could detect it. We didn’t have that technology, but now we have that technology. You should definitely use it. This is going to save a lot of lives. If you could reduce misdiagnosis and bring every doctor to a super-doctor level, I think that would be a really good thing.

Dr. Rhonda Patrick: What you’re saying is based on the current data that doctors have available to them, whether it’s an MRI, an ultrasound, or blood biomarkers. This sort of data is given to a model like GPT-5.5 Pro, for example, and with that data it can better diagnose, predict, and see things. You mentioned cancer. Is it better than a radiologist? What kind of data is implemented?

Dr. Derya Unutmaz: These are studies. I think Google did a recent study. In fact, a Science paper came out recently that used the o1-preview model, which is a very old model. The current models are probably 10 times better, or maybe more.

Dr. Rhonda Patrick: Was that almost like the first Pro model?

Dr. Derya Unutmaz: Yeah. It was the first sort of reasoning model that I tested early. It came out in September 2024. They found that the o1 model did significantly better than the average doctor in diagnosing. It didn’t miss. Imagine how good the current models are.

But I think it’s not just about diagnosing a disease, because that’s actually a small part of a doctor’s job. There’s a continuum. If you have the flu or a bacterial infection, you know what to do. You give a treatment and then you see an outcome. But that may not be true in many diseases, or even in that condition, because you might have a mutant virus or bacterium. You might have to change the treatment, or there might be a side effect.

There’s a lot of continuity there. AI can be involved in all of that process. You can continuously feed it data: “The patient received this treatment. It’s doing well. The blood pressure is down, but the patient has this symptom. What should we do? Change the dose of the drug, add this, remove that drug, or give another antibiotic?” There’s a constant process, but it isn’t always constant because people don’t go to the doctor every day. You get a prescription, see whether something works, and then go back.

What if something continuously monitored you after cancer treatment? That’s very important because cancer is a very dynamic disease. Cancer is constantly trying to survive, mutate, and counteract the immune system. You give a drug, chemotherapy works, and then the cancer comes back. Why? Mutations are accumulating. Can we catch that earlier? Can we change those decisions? Can we use multiple drugs or different drugs so that, before the cancer has an opportunity to come back, we prevent that possibility? All these decisions can be made together with AI. I think it’s going to have a tremendous impact on healthcare.

Dr. Rhonda Patrick: I do want to get back to the cancer equation in a minute. Before that, I think that not all physicians know how to use AI. They don’t know which models to use. Do they use GPT-5.5 Pro, Claude, or something else? How do they responsibly use it without outsourcing their clinical judgment? Do you have opinions on the different models to use? I know you have a collaboration with OpenAI. You’ve been one of the first scientists really testing these models in a biological arena. I think physicians listening want to know which models they should use. If we’re talking about OpenAI, it’s got to be Pro, right? It’s got to be the reasoning model. But what about Claude? What about Gemini?

Dr. Derya Unutmaz: I think people have a misunderstanding. They think of AI as, “We have the internet, so let’s just use the internet. We have AI, so let’s use AI.” But this is advancing so rapidly. The AI model we used one month ago is not the same AI model we use now. It’s doubling in intelligence every few months. I gave the example of o1-preview. Some people got stuck at the GPT-4o model. They say, “I used it and it hallucinated a lot. Even o1 wasn’t so good. It was making mistakes.” That’s ancient history.

Dr. Rhonda Patrick: That’s why I haven’t even asked about hallucinations.

Dr. Derya Unutmaz: The advantage I have is that, because I’m all in on AI, I’m continuously testing it. I can see the evolution of these models. They get 90 percent better, 95 percent better, 97 percent better. It just continuously updates itself. Right now, with the 5.5 model, I don’t see any hallucinations whatsoever. There might be 0.1 percent, but it’s extremely rare, so your trust level goes up.

It’s similar to self-driving cars. We’ve had self-driving cars for almost a decade, maybe, and they keep getting better because their AI models are updated. My advice is that doctors should not see this as something optional. They have to update their medical knowledge periodically. In fact, they have to take tests to be certified, and they have to learn about new drugs. You can’t rely only on a drug that came out five or 10 years ago. You need to know what was approved last month and update your knowledge. In a similar way, and even more so, doctors have to constantly update their AI knowledge. AI has to be part of their practice.

My recommendation is always to use the latest top model you can use. Right now, it’s GPT-5.5. For complex problems, I would always use the Pro model because it thinks for minutes. If you’re using it daily in a rapid fashion, at least use the thinking model. The thinking model is different from the instant model. The instant model is also getting better, but it needs to reason and think. Especially if you’re entering and analyzing a lot of patient data, you definitely need the Pro model.

Then there are companies like OpenEvidence. I think most doctors are starting to use it. OpenEvidence applies the latest model and keeps it updated, so doctors don’t have to worry about it. I think more companies will provide that service, so the doctor doesn’t have to worry about whether to use GPT-5.5, Claude Opus 4.7, or something else. The harness model will pick the best one for medicine and apply it.

Hospitals should implement AI just like big technology companies do. Enterprise-level AI can be more secure and protect patient data. In the hospital, you see monitors in front of the patient, such as the heartbeat monitor. There should be an AI monitor constantly monitoring the data and giving information to nurses and doctors.

Now, with AI agents, you can do that. I do it in my daily life with my email. My agents automatically check my email and tell me what’s important, so I don’t have to go through hundreds of emails. They might say, “This is waiting for you. You have a podcast with Rhonda today, so you’d better be prepared for that.” AI needs to be fully integrated, almost like a co-physician. You have AI doctors working together with real doctors.