ChatGPT

ChatGPT Health debuted in January to provide a generative AI companion for all health needs, but researchers have recently conducted a study to test its capabilities and reliability.

While there are several good qualities to the special version of the chatbot, the researchers found in their study that it downplayed various medical concerns and emergencies that should immediately warrant a trip to the emergency room.

ChatGPT Health: Study Finds It 'Underestimating' Concerns

A study conducted by researchers from the Icahn School of Medicine at Mount Sinai claims that they have conducted the first independent safety evaluation of ChatGPT Health, OpenAI's specialized chatbot for medical concerns.

Here, the researchers found that ChatGPT Health has "underestimated" several medical concerns and possible emergencies from questions or scenarios raised by the researchers. In the study, the researchers assessed the capabilities of the chatbot to perform "triage," the way medical professionals assess the patient, their state, and what they feel.

According to Gizmodo, instead of ChatGPT Health immediately directing patients to the emergency room, the chatbot suggested that they monitor their condition first for around one to two days. This happened for concerns like diabetic ketoacidosis and impending respiratory failure.

According to the study, this is despite ChatGPT Health already identifying the symptoms as early warning signs, particularly in the case of respiratory failure.

That said, for "textbook emergencies," the chatbot did more triage. However, they also noted that ChatGPT Health failed in situations where it matters most.

Is ChatGPT Health Reliable For Concerns, Emergencies?

Previous studies have been done to test ChatGPT's knowledge in the medical field, and it was found that it was not perfect in its capacity to provide advice or information.

While the chatbot may be able to answer your queries about a disease, illness, or condition, the recent study from the Icahn School of Medicine only showed that it is not yet fully reliable.

ChatGPT Health still has a lot to improve on, with studies like these looking to help improve its systems and raise awareness among the public.

Originally published on Tech Times