AI is the hottest term in health technology at CES 2025. AI chatbots, AI algorithms, and AI health suggestions can be found everywhere. However, AI has a reputation for fabricating information, and the stakes for privacy and accuracy are enormous when it comes to health.
Because of this, Movano, a company that makes smart rings, wants to be very clear about its new chatbot, EvieAI: it has only been post-trained on peer-reviewed medical articles.
The purpose of EvieAI was to provide a more accurate substitute for ChatGPT. The distinction is that, in contrast to ChatGPT and other comparable generative AI assistants, EvieAI is not expected to draw on extensive public data repositories where false information about health and wellbeing is common.
It has been trained on and will be limited to more than 100,000 medical papers authored by medical specialists, according to John Mastrototaro, CEO of Movano. According to Mastrototaro, all of the data that the LLM has access to originates from reputable sources that have been recommended by a medical advisory board. This covers articles, techniques, and procedures that have FDA approval.
Being a bounded LLM, EvieAI can only handle data from the “post-training” stage following initial creation. In this instance, it refers to medical information. The information is then compared to data from institutions such as UCLA, Harvard, and the Mayo Clinic. Before responding, the LLM consults this external data to ensure there are no conflicts. Movano claims that the outcome is 99 percent accuracy, even though we were unable to test EvieAI ourselves prior to CES. According to the business, this is feasible because whenever you ask EvieAI a question, the LLM keeps track of whether the data it has been trained on and the information provided in the discussion are correct and consistent.
It is an ambitious goal to get that degree of accuracy.