Forgot your password?
Close
typodupeerror
AI

AI Nutrition Tracking Stinks (theverge.com) 33

Posted by msmash from the thanks-for-nothing dept.
AI nutrition tracking features in popular fitness apps are producing wildly inaccurate calorie and macro counts despite promises to simplify food logging through automated photo analysis. The Verge tested AI-powered nutrition tools in Ladder, Oura Advisor, January and MyFitnessPal. Ladder's AI estimated the outlet's carefully measured 355-calorie breakfast at 780 calories and got the macro breakdown wrong even after the reviewer manually edited entries to include exact brands and amounts.

Oura Advisor routinely mistook matcha protein shakes for green smoothies. January misidentified barbecue sauce as teriyaki sauce and failed to detect mushrooms in a chicken dish. None of the apps could identify healthier ingredient swaps or accurately log ethnic foods. Oura classified a mix of edamame, quinoa and brown rice as mashed potatoes and white rice. Ladder logged dal makhani curry as chicken soup. The AI features require extensive manual corrections that negate any time savings from automated logging, the publication concluded in its scathing review.
This discussion has been archived. No new comments can be posted.

AI Nutrition Tracking Stinks More

AI Nutrition Tracking Stinks

Comments Filter:

  • Unsurprising (Score:5, Insightful)

    by necro81 ( 917438 ) on Friday November 21, 2025 @01:35PM (#65810167) Journal
    Hell, I've been cooking and eating for decades, and even I can't identify stuff on my plate sometimes, let alone estimate its mass (or volume), let alone its nutritional content. That an app using an image or two from a cellphone camera screws us doesn't surprise me.

    I could tut-tut and say these companies shouldn't be trotting out half-baked products and claiming they're better than they are. But, again, I've been around not enough to not be surprised by that, either.

    The only surprising thing is that anyone expected that these things actually would work.
    • Tut tut. I am not surprised anyone would say this if they have a very narrow exposure to human beings, and no children. Anyone who has children, or knows somebody who has children, would not be surprised that anyone expected that these things actually would work. In particular, your original impulse is spot on. You really should tut tut and say that these companies are predatory and lying about the capabilities of their products.

  • As expected (Score:3)

    by MpVpRb ( 1423381 ) on Friday November 21, 2025 @01:42PM (#65810181)

    AI is an immature tool and early adopters need to carefully cross-check the results
    Lying salesweasels may tell fantastic stories, but immature tech is still immature

  • This is all so embarrassing (Score:4, Insightful)

    by abulafia ( 7826 ) on Friday November 21, 2025 @01:44PM (#65810185)
    What happened to a sense of professionalism? Hell, of wanting people to not laugh at your work? They expect people to pay money for this junk?

    I have to imagine the C-suite at a lot of places feels intense pressure to have "an AI roadmap".

    Fine. That doesn't mean you ship not even half-baked bullshit.

    This isn't about jerking off to robot porn, it is about performance monitoring - the only thing that functionally matters is the accuracy.

    It amazes me that so much simply non-functional junk is passed off as wizzbang magic technology.

    • Re: (Score:2)

      by MobyDisk ( 75490 )

      Yes!! And:

      ...got the macro breakdown wrong even after the reviewer manually edited entries to include exact brands and amounts...

      The fact that the user hand-edited the inputs and it *still* got it wrong, tells us that the company doesn't actually care if their product even works.

    • Hey, AI is sexy, it sells, and it does not even have to be accurate. It's a dream product for any business.

  • ... this could come in handy for school kids on their cell phones.

    On mystery meat lunch days.

  • LLMs useless for everything? except scamming people and wasting bandwidth? you're telling me for the first time.
    I'm sure we will slowly go back through the backlog of news articles about 'llms being used to cure cancer!!!!' and realize that they don't actually help at all and make progress worse.
    Then finally we can all agree to stop saying 'I think LLMs are useful for some things but...'

    • What isn't a scam? How accurate are the food labels? Is 100% juice really true or a hallucination?

      • Who cares? The whole "superfood" hype is 100% scam, anyway. Scamming the gullible fools who believe in this utter junk sciemce with uselsess "AI" that some con artists vibe coded is just consequent Incredible how easily people can be scared into charlatanery. That ever same old scam has been working for millenia. Just make up some new flowery words and meaningless promises from time to time and on it goes.

      • Re: (Score:2)

        by fgouget ( 925644 )

        How accurate are the food labels?

        They'd better be otherwise the company leaves itself open to a lawsuit for false advertising. Beware of the line between a healthy amount of skepticism and nihilism.

  • Miscalculating such essential values as the precise mumbojumbo of these valuable and useful high-end alchemistic health products of the new age of quack science. People might face the most horrible consequences. Like not becoming pretty, successful, rich and immune against deseases, curses, poisonous etherand evil eye! I think we're all going to die! So horrible.

  • This is impossible for an AI or even a human to do. An AI will never be able to tell just from a photo if you glass has Coke or Diet Coke, if your dressing is full-fat or light, if that soup was made with cream or milk, or if your cupcake is made from fortified flour or a gluten free, unfortified alternative. Thinking it can is just a pipe-dream. But AI is hot so got to put AI in everything. I saw a post with an AI ready screen protector for a phone the other day. WTF?

    What they could do is have AI identify

  • I've got my own app for food tracking. Took my about 3 months of spare evenings to churn out all the code but now I've got a tracker that works exactly how I want it to work, is AI-free, accurate, free and never changes against my will.

    It's not my first app like this either.

    I've got a money tracker directly connected to my account. I've got a bass guitar learning/practicing tool. One for improving my Rubics skills. Another one for bidirectional file syncing between my desktops and mobiles. One for encrypted

    • Re: (Score:2)

      by fgouget ( 925644 )

      I've got my own app for food tracking. Took my about 3 months of spare evenings to churn out all the code

      I have one too [openfoodfacts.org] and it did not require 3 months of coding. Plus every improvement I make helps thousands. That's the power of open-source...

      • Screw open source. It's no more than a way to enable corporations to semi-legally steal someone's intellectual property and effort for nothing in exchange.

  • If even an experience chef could differentiate one curry from another WITHOUT smelling it, I would be very impressed.

    • Exactly .. the whole premise that you can accurately guess calories from a photo of a plate of food is bogus.

      Of course in simple cases of easily recognized food, that don't vary too much, you may be able to come close, but even for something like a hotdog and fries, how big is the dog, how heavy is the portion of fries?

      How is a photo going to tell you whether those carrots are just loaded with butter, or also a massive amount of sugar (an Anthony Bourdain pro-tip for delicious restaurant type carrots).

      Just

  • Should have used OpenFoodFacts (Score:3)

    by fgouget ( 925644 ) on Friday November 21, 2025 @05:33PM (#65810721)

    Then they would have gotten the down low on all the food!

    Just kidding. Still, here comes the shameless plug for OpenFoodFacts [openfoodfacts.org]:

    OpenFoodFacts is the international, crowd-sourced, open-source food database with ingredients, additives, allergens, nutrition facts (macro and micro nutrients), packaging and every bit of data one can find on the packaging. It has over 4 million products and you can use this data to compute anything you want from nutrition aspects like Nutriscore [openfoodfacts.org], ultra-transformation level (Nova [openfoodfacts.org]), to environmental impact [openfoodfacts.org] or more. Using the app you can get this data on most products with a barcode. And if none of these scores are to your liking you cant roll out your own algorithm with the criterion you care about. Besides the obvious individual use, OpenFoodFacts is also used by nutrition research institutes [openfoodfacts.org] to help their cohorts track what they eat and thus help figure out the health impact of each ingredient / additive.

    The obvious easy way to contribute [openfoodfacts.org] is to install the app [openfoodfacts.org] and add data for all the products you buy, particularly for regional products that not everyone can contribute to. You can also improve the data through the Hunger Games [openfoodfacts.org] (add ingredients, nutrition facts, hunt nutrition labels). But if you're a developer [openfoodfacts.org] your help would also be very welcome.

    And there's also Pet [openpetfoodfacts.org], Beauty [openbeautyfacts.org] and Products [openproductsfacts.org] variants if food is not your thing.

  • If AI has never eaten a grilled cheese sandwich, how would you expect it to know what a protein shake is?

  • Even if such apps would precisely recognize the type of food and its mass accurate down to 1g: The energy content of food according to literature already is inaccurate by several percent, and that would be enough to add or subtract kilograms of body mass in a year. The only reasonably precise way to measure whether you eat more or less energy than you need is by stepping on some scales from time to time, and compare the results. Luckily, the human body is pretty good at accounting for the energy it needs, e

Slashdot Top Deals

Real computer scientists don't program in assembler. They don't write in anything less portable than a number two pencil.

Close