Sam Altman, Arianna Huffington launch personalised AI health coach

OpenAI CEO Sam Altman and Arianna Huffington, author and founder of behaviour tech company Thrive Global, have launched a customised, hyper-personalised AI health coach, Thrive AI Health, to improve daily habits and chronic health issues.

Thrive AI Health will be available as a mobile app and will be integrated with Thrive Global's enterprise products. The company will train the model through peer-reviewed science, along with personal biometric and medical data that a user shares with the company.

The model will learn preferences and patterns across particular behaviours like quality of sleep, food preferences, exercise routines and stress reduction processes.


Altman and Huffington, in an online post, said that everyday behaviour contributes highly to curing chronic diseases, that warrant a majority of the US' annual health care spend.

For instance, a person with diabetes will be reminded by the AI coach to take medication on time. It will suggest meal options, or short exercises that can be done. The post goes on to say that currently, most recommendations suggested by the portal may be generic.

"Since chronic diseases—like diabetes and cardiovascular disease—are distributed unequally across demographics, a hyper-personalized AI health coach would help make healthy behaviour changes easier and more accessible," the post read.

This initiative comes amid widespread concerns about how AI chatbots and large language models use personal data to train their models. Thrive AI Health has specified that it will only use data that the user chooses to share with the app, but the data will ultimately train the app and make it more catered to the user's behaviour.

On July 9, ET took a deep dive into AI hallucinations, the biggest challenges for generative AI technology. Hallucination is a phenomenon where an AI chatbot/service offers factually incorrect information in response to a question. According to a September 2023 research report titled ‘AutoHall: Automated Hallucination Dataset Generation for Large Language Models’, the LLMs hallucination rate stood at 20-30%. This could be especially concerning in the matters of health.