🔥Today’s healthcare espresso shot is served… hot, and it’s worth reading!
🧠🌐Vol. 6 Artificial Intelligence in Medicine
🧐👀At the end of 2022, through Twitter (X) and personal discussions, I became familiar with ChatGPT, the first Artificial Intelligence (AI) program. Its first version certainly had shockingly pioneering capabilities for our technology at that time, but it was actually a language model that combined sequences of words and information in order to provide complex answers based on its massive database. Truly an amazing innovation with impressive capabilities, but also an increased risk of bias.
🚩💻Since then, a revolution has literally taken place. AI programs are evolving at a rapid pace, continuously releasing improved model versions, while more and more players are entering the field. At the same time, we have already entered the era of specialized AI programs, with models developed for conversation and information exchange (ChatGPT, Claude, etc.), text analysis and processing (ChatPDF, etc.), programming code (Cursor, etc.), video creation (RECCLOUD, etc.), design (CANVA, etc.), research (Perplexity, etc.), as well as other sectors.
🤔But what is the role of AI programs in Medical Science?
✅At present, physicians can certainly use specific programs to easily carry out their research work, mainly in writing and data analysis.
✅In parallel, specialized AI programs have been developed to analyze medical clinical scenarios and can be used as supportive tools by physicians in their clinical practice.
✅‼️Additionally, AI is expected to assist in the diagnosis and management of many diseases in the coming years. Of course, like any diagnostic test or intervention, any AI application must first be evaluated through specific predetermined protocols in order to be incorporated into current practices and find its place in diagnostic and therapeutic algorithms.
✅⚠️Finally, physicians are already using several AI programs “informally” in whatever way they feel they can benefit, and this trend is expected to continue increasing. Naturally, the same is true for people who are not physicians, whether as patients or simply for information. However, here arises a potential issue that requires attention.
🔐🆘As already mentioned, AI programs can often provide incorrect information. Furthermore, their truly astonishing capabilities are not easy to unlock. Knowledge, experience and skill are required. The information we usually provide with a simple question results in an answer that uses only 10–20% of the AI's capabilities and increases the chance of error. Therefore, proper questions, complete data and key information are needed.
🦾🧑⚕️A characteristic example from medicine is that we often ask AI about a symptom or an abnormal test result and expect it to guide us and solve the problem. However, we neglect to mention a multitude of necessary data—something expected from someone without medical training. Examples include the person’s overall profile (e.g., age, sex), full medical history, medication regimen, information from recent days that may be relevant but impossible for us to correlate, as well as seemingly unrelated accompanying symptoms.
😎💯🫵In conclusion, here are some tips I personally use when working with AI programs to maximize the validity of information, utilize a greater portion of the program’s capabilities, and minimize risk of bias:
1️⃣Act and respond as… (specific role)
2️⃣Respond at three ability levels: beginner, intermediate, expert
3️⃣Answer with bullets and checklists
4️⃣Before answering, ask me 5 clarifying questions
5️⃣Describe non-conventional solutions to the problem
6️⃣Compare advantages and disadvantages and suggest the best solution
7️⃣Analyze your answer, improve its weaknesses, and answer again
❌🔰I emphasize that I am not recommending the use of AI for health issues and the above do not constitute encouragement. But since I understand that many people use it, at least let them do it a bit more correctly.