With the emergence of artificial intelligence tools like ChatGPT, many people wonder if these technologies are good for clarifying health questions . The answer isn't simple, but these new tools have their place. First of all, you need to know how they work. Then, you must use them safely and responsibly. Artificial intelligence, LLM, ChatGPT, and others Large language models (LLMs) are artificial intelligence systems trained with enormous amounts of information , such as texts from books, scientific articles, and web pages. ChatGPT is one of the best-known examples, but there are many others, such as Gemini, Claude, or DeepSeek, and other They can produce text that appears to be written by humans, they answer questions and maintain conversations in a surprisingly natural way. They understand the context of what is asked and produce coherent responses on almost any topic. These models are constantly updated, and the latest versions have significantly improved in accuracy and ability to follow specific instructions. Their rapid evolution has caught the attention of the public and healthcare professionals, generating both enthusiasm and concern about their potential impact on medicine and heath care. What is the difference between an LLM, such as ChatGPT, and a Google search? Google is a search engine. When you search on Google, you receive a list of links to web pages where you can find information related to your search. Excluding paid advertising, the order in which these links appear is related to various factors, such as the quality and clarity of the information, the design of the pages, and many other factors that lend greater or lesser "authority" to the page. It is then up to each person to choose which pages to read and assess their credibility. It is a more laborious process, but it allows you to know the source of the information (for example, verifying whether they are websites of scientific societies, journals, or just isolated pages or promoting a specific product). More recently, search engines have also begun to present, for many searched topics, an initial summary produced by an LLM. An LLM, like ChatGPT, works differently. It constructs and presents a direct response to what is "asked," as if you were having a conversation with someone knowledgeable about a subject. You can ask follow-up questions and request clarification of answers. However, this convenience comes at a price: the model can mix information from multiple sources without citing them and sometimes generates incorrect information that seems plausible, what experts call "hallucinations." And, often, those who are asking the LLM questions are not knowledgeable about the subject and can be misled because they lack critical judgment. The choice between searching with a search engine and an LLM depends on your objective. To verify specific facts or find original sources, search engines like Google remain essential. For general explanations or understanding concepts, an LLM can be a good starting point as long as one is critical of the answers presented. Can I use LLMs to answer questions about my health? These tools are useful for: Understanding complex medical terms Learning about diseases Gaining information about healthy lifestyles Preparing questions to ask your doctor However, LLMs should never be used for: Self-diagnosis Deciding on treatments Changing or stopping medication Replacing a doctor's appointment If you have symptoms that concern you or specific questions about your health condition, always seek the advice of a qualified professional. LLMs are information tools; they are not a substitute for individualized, expert medical evaluation. Are there risks in using an LLM to make health decisions? Yes, there are, for example, basing these decisions on: Incorrect information. LLMs can generate answers that appear correct but are not. In fact, their responses indicate that the answer may contain errors. Outdated information. Medical information evolves rapidly. A model trained months ago may not be considering more recent discoveries or changes in clinical guidelines. Biased information. The data used in training may not consider information related to gender, ethnicity, socioeconomic status, or other factors, leading to less appropriate answers for certain groups. Furthermore: The articulate and confident way LLMs respond to questions and searches can lead us to over-confidence in their answers, even when they are incorrect, leading to false confidence . Sharing personal medical information with these tools can compromise your privacy . Relying solely on these tools can worsen health problems by delaying seeking medical care . Therefore, the correct advice regarding the use of LLMs for healthcare is: Use these tools as a complement, never as a substitute for medical advice. Share the information you find and any questions it raises with your doctor. Use LLMs, such as ChatGPT, but always securely Protect your privacy: Never enter personal information such as your full name, health number, or detailed medical history. Be specific in your questions: The more context you provide (without personal data), the better the answers will be. Question and verify: Ask the LLM to explain any uncertainties or limitations in their answers. Always confirm important information with reliable sources. Ask the LLM for information sources with links that you can validate. Use the LLM as a starting point: Consider the responses as initial information to later discuss with your doctor. Recognize warning signs: For severe, sudden, or worrisome symptoms, seek medical attention immediately, not the LLM. What will the future bring More accurate models are expected in the future, capable of recognizing when they don't know something and indicating the degree of confidence in their answers. Specialized versions will emerge in various fields, including medicine, validated and certified for different uses (by doctors, by patients, for example), with greater safety. The success of this transformation will depend on collaboration between technicians, healthcare professionals, regulators, and patients. The goal is not to replace the doctor-patient relationship, but to enrich it with tools that improve access to information and the quality of healthcare. With these tools, we may have the capacity to have super doctors and super patients! In short… LLMs, of which ChatGPT is one of the best known, are a promising tool, but they have limitations . For now, they can help us better understand our health and prepare for more productive conversations with healthcare professionals. Embrace new technologies but always maintain a critical eye . Your health deserves the best possible care, and that means combining the best technology with the experience and human judgment of your doctor and other healthcare professionals. Author : José Ferreira Santos (cardiology) Sources Clusmann J, Kolbinger FR, Muti HS, et al. The future landscape of large language models in medicine . Commun Med 2023;3:141. Lee P, Bubeck S, Petro J. Benefits, limits, and risks of GPT-4 as an AI Chatbot for medicine . N Engl J Med 2023;388:1233-9. Singhal K, Azizi S, Tu T, et al. Large language models encode clinical knowledge. Nature 2023;620:172-80. The content published in the Health Guide is intended for the general public and is provided for informational and educational purposes only. It does not replace individual medical assessment, diagnosis, or personalized clinical advice from a qualified healthcare professional. The content is produced or reviewed by physicians and other healthcare professionals at Hospital da Luz and is based on the knowledge available at the time of publication or the most recent update, as indicated at the beginning of each item.
