Exploring AI in Nutrition Tracking: Are We There Yet?
This is Optimizer, a weekly newsletter sent every Friday from Verge senior reviewer Victoria Song that dissects and discusses the latest phones, smartwatches, apps, and other gizmos that swear they’re going to change your life. Optimizer arrives in our subscribers’ inboxes at 10AM ET. Opt in for Optimizer here.
The AI Promise and Reality
Once again, AI is failing to deliver on some of its promises, especially regarding nutrition tracking. Before my last long run, I made my customary pre-workout breakfast: two dark chocolate Kodiak protein waffles, a tablespoon of peanut butter, and a drizzle of honey. On the side, a modest cup of iced coffee with a splash of soy milk.
As someone who dabbles with counting macros, I outlined this breakfast’s nutritional info: roughly 355 calories, 16g of protein, 28g of carbs, and 17g of fat. I’ve spent five training blocks fine-tuning these details to optimize my performance, but I find the tedious task of reentering this data into various logging apps cumbersome.
AI and the Quest for Effortless Logging
Enter AI, which aims to revolutionize food tracking. Recently, my strength training app, Ladder, introduced AI-powered nutrition features promising to simplify macro counting. Just take a picture, and the technology would do the rest. To my dismay, the AI estimated my carefully curated breakfast at a staggering 780 calories, 20g protein, 92g carbs, and 39g fat. Even when I specifically edited the details to include precise brands and amounts, it produced another, similarly incorrect estimate. This experience cemented my decision to abandon calorie and macro counting.
Why Food Logging is So Challenging
Food logging is notoriously arduous. Traditional apps can help you search for food options, but real-life situations often complicate accuracy. Dining out presents challenges, as many restaurants may share calorie counts without macro breakdowns. Even home-cooked meals become unclear; experienced cooks often improvise recipes and switch ingredients on the fly. Logging accurately requires measuring every single item eaten, which can feel restrictive.
AI can identify broadly what’s in the picture, but gets specifics wrong. Screenshot: Ladder
AI Innovations in Health Tracking
Despite these challenges, AI tools keep emerging. Oura has incorporated an Advisor chatbot capable of analyzing meals through written descriptions or photos, offering insights about nutritional value and potential health impacts. Integration with Dexcom continuous glucose monitors allows users to see real-time changes linked to meal intake, enhancing the awareness and feedback loop.
This bowl of pasta has kale, peppers, mushrooms, and ground turkey. I see no cheese puffs or ice cream. Screenshot: Oura
Similarly, the January app allows for meal photos, generating an estimate of glucose level impacts based on individual data. MyFitnessPal now features a ScanMeal function for quick logging through photos, while emerging gamified food-tracking apps use interactive elements, like AI raccoon pets, to engage users.
The Limitations of Current AI Technology
Despite these advancements, AI’s food identification struggles remain a significant hurdle. Applications routinely misclassify foods: my matcha protein shake becomes a green smoothie, while certain meals are mistaken for entirely different dishes. The AI errors extend beyond recognition to estimations of portion sizes, often leading to gross inaccuracies when attempting to log meals accurately.
These tools may allow users to edit AI-generated entries, but this ultimately negates the simplicity that AI is meant to provide. Spending time correcting inaccuracies feels more like replacing one tedious task with another.
Reassessing the Problem to Solve
Upon reflection, leveraging AI to simplify food logging might be misdirected. Dietary changes face complexities not solely due to knowledge gaps but also emotional and behavioral challenges ingrained in our patterns. AI can prompt us toward better choices, but real change requires effort on our part.
The true goal of food logging transcends arbitrary calorie counts. It’s about cultivating awareness of our eating habits, identifying areas for improvement, and learning to enjoy food mindfully—without forgetting the fun in indulgence. Ideally, competency in understanding one’s dietary preferences leads to a point where logging isn’t necessary.
In essence, a successful app fosters independence over reliance, empowering users to trust their decisions rather than feeling shackled to technology.
Not pictured: Me editing this to be a matcha latte, not a kale smoothie, and clarifying there are apples in this tart. Photo by Amelia Holowaty Krales / The Verge
Ultimately, a successful food logging application needs to engage users without fostering dependence. AI can offer insights and facilitate tracking, but human agency remains paramount in achieving sustainable health habits. Instead of being an auxiliary tool, perhaps AI should serve as a partner, encouraging users to reflect on their dietary choices proactively.
Explore the journey of AI in nutrition tracking and consider where it might lead us next. For more information, visit the original article here.
Image Credit: www.theverge.com
