An nameless reader quotes a report from Ars Technica: On Monday, Microsoft and Epic Techniques introduced that they’re bringing OpenAI’s GPT-4 AI language mannequin into well being care to be used in drafting message responses from well being care staff to sufferers and to be used in analyzing medical data whereas in search of tendencies. Epic Techniques is one in all America’s largest well being care software program firms. Its digital well being data (EHR) software program (reminiscent of MyChart) is reportedly utilized in over 29 % of acute hospitals in the USA, and over 305 million sufferers have an digital file in Epic worldwide. Tangentially, Epic’s historical past of utilizing predictive algorithms in well being care has attracted some criticism up to now.
In Monday’s announcement, Microsoft mentions two particular methods Epic will use its Azure OpenAI Service, which offers API entry to OpenAI’s massive language fashions (LLMs), reminiscent of GPT-3 and GPT-4. In layperson’s phrases, it implies that firms can rent Microsoft to offer generative AI providers for them utilizing Microsoft’s Azure cloud platform. The primary use of GPT-4 comes within the type of permitting docs and well being care staff to routinely draft message responses to sufferers. The press launch quotes Chero Goswami, chief data officer at UW Well being in Wisconsin, as saying, “Integrating generative AI into a few of our day by day workflows will improve productiveness for a lot of of our suppliers, permitting them to concentrate on the scientific duties that really require their consideration.” The second use will convey pure language queries and “knowledge evaluation” to SlicerDicer, which is Epic’s data-exploration device that enables searches throughout massive numbers of sufferers to establish tendencies that could possibly be helpful for making new discoveries or for monetary causes. In keeping with Microsoft, that may assist “scientific leaders discover knowledge in a conversational and intuitive method.” Think about speaking to a chatbot much like ChatGPT and asking it questions on tendencies in affected person medical data, and also you would possibly get the image. Dr. Margaret Mitchell, chief ethics scientist at Hugging Face, is anxious about GPT-4’s capability to make up data that is not represented in its knowledge set. One other concern is the potential bias in GPT-4 that may discriminate in opposition to sure sufferers based mostly on gender, race, age, or different elements.
“Mixed with the well-known downside of automation bias, the place even specialists will imagine issues which might be incorrect in the event that they’re generated routinely by a system, this work will foreseeably generate false data,” says Mitchell. “Within the scientific setting, this could imply the distinction between life and loss of life.”
