Google Gemini proves to be a better health coach than humans

It's time to celebrate the incredible women leading the way in AI. Nominate your incredible leader for the VentureBeats Women in AI Awards by June 18. Learn more here

Google Gemini has only been around for six months, but it's already showing impressive capabilities in areas such as security, coding, and debugging (though of course, it also has significant limitations).

Large-scale language models (LLMs) are now outperforming humans when it comes to giving sleep and fitness advice.

Google researchers have presented the Personal Health Large Language Model (PH-LLM), a version of Gemini fine-tuned to understand and reason about time-series personal health data from wearable devices such as smartwatches and heart rate monitors. In experiments, the model answered questions and made predictions significantly better than experts with years of experience in the health and fitness field.

Breaking News: Google releases large-scale language model for personal health

Fine-tune with Gemini Read your wearable data and find personalized insights and recommendations Outperformed professional sleep and fitness experts in certification exams pic.twitter.com/FjXcYTIGON

— Chef Jeff (@chefjeffsf) June 11, 2024

“Our study employs generative AI to extend the utility of our models beyond simply predicting health states to also provide coherent, contextual and potentially normative outputs in response to complex health behaviors,” the researchers wrote.

VB Transform 2024 registration opens

Join enterprise leaders at our flagship AI event in San Francisco July 9-11. Network with your peers, explore the opportunities and challenges of generative AI, and learn how to integrate AI applications in your industry. Register now

Gemini as sleep and fitness expert

Wearable technology can help people monitor and ideally meaningfully improve their health. These devices passively and continuously capture inputs such as exercise and food logs, mood diaries, and sometimes social media activity, providing a rich, long-term source of data for monitoring an individual's health, the Google researchers noted.

However, data collected on sleep, physical activity, cardiometabolic health, and stress is rarely integrated into clinical practice due to its sporadic nature. Researchers believe this is because the data is collected without context, requires extensive computation to store and analyze, and can be difficult to interpret.

And while JDs perform well when it comes to answering medical questions, analyzing electronic medical records, making diagnoses based on medical images, and psychiatric evaluations, they often lack the ability to reason and make recommendations based on data from wearables.

But Google researchers have made breakthroughs in training PH-LLM to make recommendations, answer specialized exam questions, and predict outcomes of self-reported sleep problems and sleep disorders. The model was given multiple-choice questions, and the researchers also performed thought chaining (mimicking human reasoning) and zero-shot techniques (to recognize objects or concepts that had not been encountered before).

Remarkably, the PH-LLM achieved an average score of 79% on the sleep test and 88% on the physical fitness test, both of which exceeded the average scores of a human expert sample that included five professional athletic trainers (average years of experience 13.8 years) and five sleep medicine specialists (average years of experience 25 years). The humans achieved an average score of 71% on the physical fitness test and 76% on the sleep test.

In one example coaching recommendation, the researcher instructed the model: “You are a sleep medicine expert. You have been given the following sleep data. The user is male and 50 years old. Please list the most important insights.”

PH-LLM responded: They have trouble falling asleep. They don't get enough deep sleep. [is] It's important for your body to recover. The model further advises: “Keep your bedroom cool and dark, avoid naps, and maintain a consistent sleep schedule.”

On the other hand, when asked from four answer options what type of muscle contraction occurs in the pectoralis major during the slow, controlled lowering phase of the bench press, PH-LLM correctly answered eccentric.

For the income recorded by the patient, the researchers asked the model, “Based on this wearable data, would the user report having trouble falling asleep?” To which the model responded, “This person is likely to report having trouble falling asleep several times in the past month.”

The researchers note: “Although further development and evaluation is needed in the personal health domain where safety is crucial, these results demonstrate both the broad knowledge base and capabilities of the Gemini model.”

Gemini can provide personalized insight

To achieve these results, the researchers first created and curated three datasets to test personalized insights and recommendations derived from recordings of physical activity, sleep patterns, and physiological responses, as well as expert domain knowledge and predictions about self-reported sleep quality.

They worked with domain experts to create 857 case studies representing real-world scenarios around sleep and fitness: 507 for the former and 350 for the latter. The sleep scenarios use individual metrics to identify potential causes and provide tailored recommendations to improve sleep quality. The fitness tasks use information from training, sleep, health metrics, and user feedback to make recommendations on the intensity of physical activity on a given day.

Both case study categories incorporate wearable sensor data, demographic information (age and gender), and expert analysis, spanning up to 29 days for sleep and over 30 days for fitness.

Sensor data included an overall sleep score, variability in resting heart rate and heart rate variability, time asleep (start and end times), time awake (minutes), restlessness, percentage of REM sleep, breathing rate, steps taken, and fat-burning minutes.

“Our study demonstrates that PH-LLM can integrate objective data passively obtained from wearable devices into personalized insights, potential causes of observed behaviors, and recommendations to improve sleep hygiene and fitness outcomes,” the researchers wrote.

There's still a lot of work to be done in personal health apps

Still, the researchers acknowledge that PH-LLM is still in its infancy, and like any emerging technology, there are bugs to work out. For example, the responses generated by the model were not always consistent, there were notable differences in confabulation across case studies, and the LLM's responses were sometimes conservative or cautious.

In the fitness case study, the model was sensitive to overtraining and, in one instance, human experts noted that they failed to identify lack of sleep as a potential cause of harm. Also, because the case study sampled from a broad demographic and relatively active individuals, it may not be fully representative of the population and could not address a broader range of sleep and fitness issues.

The researchers caution that much work remains to be done to ensure LLM is reliable, safe, and unbiased in personal health applications, including further reducing confabulation, accounting for unique health conditions not captured by sensor information, and ensuring that training data reflects a diverse population.

But overall, the researchers note, “the results of this study represent an important step toward LLM providing personalized information and recommendations that support individuals in achieving their health goals.”

VB Daily

Stay up to date! Get the latest news every day by email

By subscribing, you agree to VentureBeat's Terms of Use.

Thanks for subscribing! Check out other VB newsletters here.

An error occurred.

Sources 1/ https://Google.com/ 2/ https://venturebeat.com/ai/google-gemini-proves-a-better-health-coach-than-humans/ The mention sources can contact us to remove/changing this article

What Are The Main Benefits Of Comparing Car Insurance Quotes Online

LOS ANGELES, CA / ACCESSWIRE / June 24, 2020, / Compare-autoinsurance.Org has launched a new blog post that presents the main benefits of comparing multiple car insurance quotes. For more info and free online quotes, please visit https://compare-autoinsurance.Org/the-advantages-of-comparing-prices-with-car-insurance-quotes-online/ The modern society has numerous technological advantages. One important advantage is the speed at which information is sent and received. With the help of the internet, the shopping habits of many persons have drastically changed. The car insurance industry hasn't remained untouched by these changes. On the internet, drivers can compare insurance prices and find out which sellers have the best offers. View photos The advantages of comparing online car insurance quotes are the following: Online quotes can be obtained from anywhere and at any time. Unlike physical insurance agencies, websites don't have a specific schedule and they are available at any time. Drivers that have busy working schedules, can compare quotes from anywhere and at any time, even at midnight. Multiple choices. Almost all insurance providers, no matter if they are well-known brands or just local insurers, have an online presence. Online quotes will allow policyholders the chance to discover multiple insurance companies and check their prices. Drivers are no longer required to get quotes from just a few known insurance companies. Also, local and regional insurers can provide lower insurance rates for the same services. Accurate insurance estimates. Online quotes can only be accurate if the customers provide accurate and real info about their car models and driving history. Lying about past driving incidents can make the price estimates to be lower, but when dealing with an insurance company lying to them is useless. Usually, insurance companies will do research about a potential customer before granting him coverage. Online quotes can be sorted easily. Although drivers are recommended to not choose a policy just based on its price, drivers can easily sort quotes by insurance price. Using brokerage websites will allow drivers to get quotes from multiple insurers, thus making the comparison faster and easier. For additional info, money-saving tips, and free car insurance quotes, visit https://compare-autoinsurance.Org/ Compare-autoinsurance.Org is an online provider of life, home, health, and auto insurance quotes. This website is unique because it does not simply stick to one kind of insurance provider, but brings the clients the best deals from many different online insurance carriers. In this way, clients have access to offers from multiple carriers all in one place: this website. On this site, customers have access to quotes for insurance plans from various agencies, such as local or nationwide agencies, brand names insurance companies, etc. "Online quotes can easily help drivers obtain better car insurance deals. All they have to do is to complete an online form with accurate and real info, then compare prices", said Russell Rabichev, Marketing Director of Internet Marketing Company. CONTACT: Company Name: Internet Marketing CompanyPerson for contact Name: Gurgu CPhone Number: (818) 359-3898Email: [email protected]: https://compare-autoinsurance.Org/ SOURCE: Compare-autoinsurance.Org View source version on accesswire.Com:https://www.Accesswire.Com/595055/What-Are-The-Main-Benefits-Of-Comparing-Car-Insurance-Quotes-Online View photos