TL;DR:
- Wearable devices have gained widespread adoption, with over 40% of US households owning one.
- AI is now advancing to create personalized AI health coaches using wearable data.
- Research shows wearables can positively impact behavior, but their effects may diminish over time.
- AI can interpret and provide insights from wearable data, potentially leading to more effective health coaching.
- Large language models like GPT are being used to analyze wearable data for mental health diagnoses.
- AI health coaches have the potential to offer deeper insights and guidance compared to human coaches.
- Tech giants like Google and Apple are exploring AI health coach initiatives.
- Challenges include maintaining AI health coach effectiveness over time and translating data into tangible health outcomes.
Main AI News:
A decade ago, monitoring your daily steps or heart rate might have seemed peculiar. Back then, proponents of the quantified self movement delivered passionate TED Talks, while journalists covered the emergence of this novel trend. Today, in the United States alone, over 40% of households possess a wearable device, as reported by the statistics service Statista. It has become commonplace to overhear retirees comparing step counts for the day. The era of the quantified self has arrived.
As artificial intelligence continues its relentless progress, researchers and technologists are now poised to take the next leap: the creation of AI health coaches capable of analyzing health data and providing users with personalized guidance on maintaining optimal health.
The Triumph of the Quantified Self
There is substantial evidence supporting the benefits of wearables. A review of scientific studies conducted in 2022, involving over 160,000 participants, revealed that those assigned to wear activity trackers took an additional 1,800 steps per day, resulting in an average weight loss of around two pounds.
Wearables influence behavior in various ways. They encourage users to set goals, enable them to monitor aspects of personal interest and provide reminders when they veer off course, according to Carol Maher, a professor of population and digital health at the University of South Australia and co-author of the review.
However, these effects tend to diminish over time, cautions Andrew Beam, an assistant professor at the Harvard T.H. Chan School of Public Health specializing in medical artificial intelligence.
Unlocking the Potential of AI
The accurate interpretation of crucial health metrics from data inputs, such as determining step count from a wrist-worn accelerometer, necessitates AI, albeit a somewhat unglamorous variety, notes Shwetak Patel, a professor in computer science and engineering at the University of Washington and director of health technologies at Google. Yet, he emphasizes that AI can expand the capabilities of these sensors in unforeseen ways, including features like fall detection and blood oxygen monitoring, which are already available on popular wearable devices. Some researchers are even striving to utilize the relatively basic health data from wearables for disease detection, including COVID-19, though generally with slightly less accuracy than clinical-grade devices.
Thus far, AI has played a supporting role in the rise of the quantified self. Researchers are eager to leverage recent advancements to propel AI into a central role.
The Emergence of AI Health Coaches
Patel co-authored a recent paper in which researchers fed data from wearables into large language models like OpenAI’s GPT series. The models then generated insights based on the data, which could prove valuable for clinicians diagnosing mental health conditions. For instance, if a participant’s sleep data exhibited irregular patterns, the AI system would flag this and suggest that erratic sleep might indicate issues like stress, anxiety, or other disorders.
According to Patel, the next generation of AI models possesses the ability to reason, making them suitable for personalized health coaching. “It’s one thing to say, ‘Your average heart rate is 70 beats per minute,’ but what we’re focusing on is how to interpret that within your unique context,” Patel explains.
Wearable data could enable AI “coaches” to gain a more profound understanding of users’ health than human coaches ever could, offering detailed, objective insights, such as sleep patterns.
Carol Maher, who co-authored a review of the research on AI chatbots’ impact on lifestyle behaviors, notes that chatbot health coaches have been found to help people increase physical activity, improve sleep quality, and enhance their diets. Although their effects are somewhat smaller than those typically observed with wearables, Maher anticipates that more sophisticated AI health coaches, like ChatGPT, could achieve greater effectiveness. However, she acknowledges challenges remain, such as the models’ tendency to generate fabricated information.
The Skeptical Perspective
Andrew Beam raises valid concerns about chatbot health coaches, highlighting that they also suffer from diminishing effectiveness over time, akin to wearables. Moreover, even human scientists, armed with extensive individual health data, struggle to provide personalized advice in the realm of health.
Nevertheless, even in the absence of precise recommendations based on health data, AI health coaches can monitor the impact of actions and adjust their guidance accordingly. For instance, heart rate data collected during a suggested workout could inform future exercise recommendations, as noted by Sandeep Waraich, product management lead for wearable devices at Google.
While Google has not officially announced plans for an AI health coach, it intends to provide AI-powered insights to Fitbit users in early 2024. Additionally, reports indicate that Apple is working on an AI health coach, codenamed Quartz, slated for release next year.
Beyond the tech giants, smaller players are also harnessing wearable data for personalized health coaching. Health app Humanity claims to determine a user’s “biological age” with a three-year accuracy margin based on movement and heart rate data. The algorithm behind Humanity’s offering, developed using data from the U.K. biobank, aims to track changes in biological age over time, helping users discern the effects of their actions on their overall health.
Nonetheless, skeptics like Andrew Beam emphasize the need for caution, as there is still insufficient evidence linking these measures to tangible health outcomes. This challenge underscores the broader concerns surrounding AI’s role in healthcare, where translating AI algorithms into tangible improvements in patient health remains a work in progress.
Conclusion:
The rise of AI health coaches signifies a transformative shift in the market. With increasing consumer adoption of wearables, the integration of AI to provide personalized health coaching presents a significant opportunity for tech companies. However, challenges related to sustaining the effectiveness of AI coaches and proving their impact on health outcomes need to be addressed for this market to realize its full potential.
