Allowing AI to give medical advice is associated with great risks, according to a new study where AI was asked to answer a number of common questions about medicines. In more than a fifth of cases, AI-powered chatbots provided incorrect information about medicines. The study showed that four out of ten AI responses could even lead to harm.
– and in some cases even death. In addition to outright inaccuracies, the answers were also very difficult to understand. Today, one in five British GPs uses programmes like ChatGPT and Bing AI at work. Experts have warned that this could lead to misdiagnoses in healthcare.
Source: British Medical Journal