In a recent experiment, the author tested three major chatbots - Google Gemini, Meta Llama 3, and ChatGPT - by asking medical questions they already knew the answers to. The AI responses were generally 80% correct and 20% incorrect, with answers changing slightly each time the question was repeated. The author found it concerning how misinformation from AI could potentially harm people, particularly those with rare medical diseases. While there is hype around using AI to improve the healthcare system for children with special needs, the author argues that the complex problems they face require more than simple solutions from AI. Increasing payment rates and providing more time for healthcare professionals to communicate with patients and families is emphasized as a necessary step. The author also discusses the limitations of AI in providing reliable medical information, as AI often synthesizes information from the internet without clear sources.

Health literacy is highlighted as a valuable skill, and the author suggests that AI can potentially provide inaccurate information, especially considering the abundance of unverified AI-generated content on the internet. The risks of AI in healthcare are emphasized, as incorrect information from AI could have severe consequences. AI's ability to connect families with services is also questioned, as it lacks the ability to provide culturally competent and personalized assistance. Additionally, the author raises concerns about AI being used by health insurance companies to make decisions on patient care, as it can perpetuate biases present in the current healthcare system. The author argues for legislation that ensures physician oversight and adequate safeguards for the use of AI in healthcare. Ultimately, the author stresses the importance of listening to the needs of those who rely on the healthcare system rather than solely relying on AI solutions.