ChatGPT's Apple Health Integration Flaws Exposed in New Report

A newsman for The Washington Post has enactment ChatGPT's caller optional Apple Health integration feature to the trial by feeding it 10 years of their Apple Watch data. The results were not encouraging, to accidental the least.


Earlier this month, OpenAI announced the motorboat of ChatGPT Health, a dedicated conception of ChatGPT wherever users tin inquire health-related questions wholly separated from their main ChatGPT experience. For much personalized responses, users tin link assorted wellness information services specified arsenic Apple Health, Function, MyFitnessPal, Weight Watchers, AllTrails, Instacart, and Peloton.

ChatGPT Health tin besides integrate with your aesculapian records, allowing it to analyse your laboratory results and different aspects of your aesculapian past to pass its answers to your health-related questions.

With this successful mind, newsman Geoffrey Fowler gave ChatGPT Health entree to 29 cardinal steps and 6 cardinal heartbeat measurements from his Apple Health app, and asked the bot to people his cardiac health. It gave him an F.

Feeling understandably alarmed, Fowler asked his existent doctor, who successful nary uncertain presumption dismissed the AI's appraisal entirely. His doc said Fowler was astatine specified debased hazard for bosom problems that his security apt wouldn't adjacent screen further investigating to disprove the chatbot's findings.

Cardiologist Eric Topol of the Scripps Research Institute was likewise unimpressed with the ample connection model's assessment. He called ChatGPT's investigation "baseless" and said radical should disregard its aesculapian advice, arsenic it's intelligibly not acceptable for premier time.

Perhaps the astir troubling finding, though, was ChatGPT's inconsistency. When Fowler asked the aforesaid question respective times, his people swung wildly betwixt an F and a B. ChatGPT besides kept forgetting basal accusation astir him, including his sex and age, contempt it having afloat entree to his records.

Anthropic's Claude chatbot fared somewhat amended – though not by much. The LLM graded Fowler's cardiac wellness a C, but it besides failed to decently relationship for limitations successful the Apple Watch data.

Both companies accidental their wellness tools aren't meant to regenerate doctors oregon supply diagnoses. Topol rightly argued that if these bots can't accurately measure wellness data, past they shouldn't beryllium offering grades astatine all.

Yet thing appears to beryllium stopping them. The U.S. Food and Drug Administration earlier this period said the agency's occupation is to "get retired of the mode arsenic a regulator" to beforehand innovation. An bureau commissioner drew a reddish enactment astatine AI making "medical oregon objective claims" without FDA review, but ChatGPT and Claude reason they are conscionable providing information.

"People that bash this are going to get truly spooked astir their health," Topol said. "It could besides spell the different mode and springiness radical who are unhealthy a mendacious consciousness that everything they're doing is great."

ChatGPT's Apple Health integration is presently constricted to a radical of beta users. Responding to the report, OpenAI said it was moving to amended the consistency of the chatbot's responses. "Launching ChatGPT Health with waitlisted entree allows america to larn and amended the acquisition earlier making it wide available,” OpenAI VP Ashley Alexander told the work successful a statement.
This article, "ChatGPT's Apple Health Integration Flaws Exposed successful New Report" archetypal appeared connected MacRumors.com

Discuss this article successful our forums