Ce médecin a construit une IA qui travaille pour lui! Dr Julien Welmant

Chapters: 0:12 Medical AI in Everyday Life 3:53 Finding the Right Bibliography 8:29 Scribia, the AI ​​That Writes 10:34 Remote Monitoring and the Future 16:56 Hallucinations and Reliable Tools 19:13 GDPR, the Real Danger 24:57 Understanding AI Agents 31:01 Atlas, the Local System 38:21 Vibe Coding in Clinical Practice 41:41 Final Thoughts The discussion focuses on the use of artificial intelligence in medicine, with an emphasis on truly useful everyday applications. Julien Welmant explains that the most prevalent AI today is generative AI, popularized by ChatGPT, but that its medical value depends heavily on the tool used and the context of its use. The first application deemed reliable is bibliographic research. He describes tools like Cliniscan, PaperDoc, and Open Evidence, used to quickly find sourced medical literature on rare or complex situations. He specifies that, in this context, the absence of patient data limits the risks. A second important use case involves transcribing consultations and generating reports, prescriptions, or internal messages. He cites solutions like Nabla, Doctolib, and other AI scribe tools, which save time, particularly during consultations, while maintaining human validation. The discussion then addresses the limitations of generic models like ChatGPT, Gemini, and Claude. Julien emphasizes the potential for misleading information, including fabricated studies, and the physician's responsibility, as they remain solely accountable for their prescriptions. He believes that medical tools should be specialized. The issue of confidentiality is discussed at length. Julien explains that even anonymized or pseudonymized records can be cross-referenced and re-identified. He therefore recommends never sending patient data to foreign services and instead favoring HDS-certified hosting or a strictly local system. Finally, he presents his Atlas system: a local Nvidia server, disconnected from the internet, on which he has assembled open-source components to create his own tools. He discusses report generation, specialized dictation, internal messaging, and task automation. He advocates for testing and co-creating these tools while remaining cautious about security and technological dependence. 👉 Subscribe to the Super Recap newsletter to receive a 1-minute email summarizing the main points of the week's episodes (it's free and spam-free!): https://superdocteurs.substack.com/ My book is available here: https://www.chroniquesociale.com/comp... The statements made in this podcast do not constitute medical advice or an inducement to prescribe medication. Not all statements made are verified. Listeners are encouraged to consult current official recommendations and supplement their education with recognized professional sources. Instagram:   / dr.matthieu.cantet   YouTube:    / @superdocteurs   LinkedIn:   / matthieu-cantet-4a5591294