Live data from Hacker News

I used AI-powered calorie counting apps, and they were even worse than expected

lifehacker.com

191–200 of 302 posts

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#191
The benefit of AI as a feature in calorie counting apps is great. But you have to track stuff manually for a while to realize the benefit. Specifically: using AI for restaurant meals is a convenient starting point, from which you use your hard-won feel for weight & caloric density of things to tweak the first pass to something closer to the mark.

As of now I would not at all trust a fully AI calorie counting app

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#192

Others have found that CalAI likely uses Gemini 1.5 (not sure which variant) which is used because it's very cheap, not for its accuracy. I understand the fundamental issues with calorie counting via photo analysis/LLM. However these articles would be more interesting if they included comparison with results from actual cutting-edge LLMs, rather than the budget models. I acknowledge that apps don't use these models t…

> I don't think it's being used in CalAI though. They don't, but that doesn't stop them from lying about it on their homepage. > Snap a photo with Cal AI, and your phone's depth sensor calculates food volume. Our AI then analyzes and breaks down your meal to determine calories, protein, carbs, and fat. The grift must go on.

Oh interesting. I wouldn't put it past them that they include that data in the input though. Why are you certain they don't? (I guess that they don't, too, because someone reported getting identical calorie readings from giving photos to Gemini 1.5 directly)

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#193
post #168

Earlier quoted context omitted.

Is it ironic that my AI detector is ringing alarm bells at this comment (both its overall cadence, as well as it’s particularly unnecessary use of an em dash)?

why are people insisting that an em-dash is exclusive to llms somehow, when on a mac it's super easy to type dash, en-dash and em-dash — just type dash, option + dash or shift + option + dash. I got used to defaulting to em-dash (even if incorrectly) and now I'll be categorized as an llm for that?

Yes, is the simple answer to your question.

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#194
post #184

Founder of https://snapcalorie.com and ex-Google AI researcher here. The article actually gets our pricing wrong, it's completely free to use SnapCalorie, the $79.99 price is a donation to support our research and help us continue providing the best tools to everyone for free. We've published in top peer reviewed academic conferences and our algorithm (can't say the same for other apps) IS actually more accurate than…

What about when people give up because you massively underestimate their calories? Or when people develop eating disorders because you massively overestimate their calories? If the goal is to learn then accurate information is important, although I suppose it's harder to get a VC to fund.

We do a lot of research on preventing eating disorders. As the article mentions we did not suggest she lose weight down to being underweight like other apps did.

Our app has the fruit and veg counters ABOVE the calorie meter on the dashboard. The reason for this is that maximization mindset (e.g. maximizing fruits and veg) is way healthier than a minimization mindset (e.g. minimizing calories or carbs).

We actually even tried to fully remove calories from the app at some point but we had a vast majority of users churn and decided it would be healthier for people who want calorie tracking to stick with our app by having it present, but requiring them to scroll past the features that promote a healthier mindset to get there. Feedback from users has been amazing that they've slowly started focusing more on fruits and veg.

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#195

The benefit of AI as a feature in calorie counting apps is great. But you have to track stuff manually for a while to realize the benefit. Specifically: using AI for restaurant meals is a convenient starting point, from which you use your hard-won feel for weight & caloric density of things to tweak the first pass to something closer to the mark. As of now I would not at all trust a fully AI calorie counting app

This is exactly correct. I find AI in my food tracking app (FoodNoms) to be really useful, but no, you cannot rely on it on its own. You always have to tweak the results to match what you feel in your gut (no pun intended) the counts should be. Interestingly, very similar situation to using AI for code: you shouldn't just blindly trust what it puts out, but if you know what you're doing, it can save you a lot of time.

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#196

The benefit of AI as a feature in calorie counting apps is great. But you have to track stuff manually for a while to realize the benefit. Specifically: using AI for restaurant meals is a convenient starting point, from which you use your hard-won feel for weight & caloric density of things to tweak the first pass to something closer to the mark. As of now I would not at all trust a fully AI calorie counting app

100%. You should not trust photo based calorie tracking and should only use it when there is no other realistic choice (e.g. eating out).

You especially should not trust an app that does not do any research on the accuracy of it's algorithm or claims to be "consistently accurate" as many of them do currently. The truth is these algorithms have the potential to be accurate when values are averaged out over the long term, but dish to dish they will have occasionally wildly inaccurate results.

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#197
post #184

Founder of https://snapcalorie.com and ex-Google AI researcher here. The article actually gets our pricing wrong, it's completely free to use SnapCalorie, the $79.99 price is a donation to support our research and help us continue providing the best tools to everyone for free. We've published in top peer reviewed academic conferences and our algorithm (can't say the same for other apps) IS actually more accurate than…

What about when people give up because you massively underestimate their calories? Or when people develop eating disorders because you massively overestimate their calories? If the goal is to learn then accurate information is important, although I suppose it's harder to get a VC to fund.

The calorie counts on food packets themselves are +/- 20%, so it's inaccurate all the way down.

Some of the value also just comes from writing down what you eat, and noting the snacks, dressings, glasses of wine etc that you forget about but all add up.

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#198
post #107

Earlier quoted context omitted.

> most people underestimate the calories in alcohol Funny, I feel the other way around. I kept hearing about how much calories there are in alcohol, and then when I started calorie counting I didn't find it so high. Like 6 shots of gin are ~550kcal and enough to get anyone pretty drunk. Unless one is a regular heavy drinker it's not that hard to once in a while budget calories during the day to be able to get a few s…

Sure six shots are 550 calories. But who drinks six shots of gin? You need to also include all the mixers as well and they're usually all sugar

For a G&T, you can get slimline tonic which tastes quite good.

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#199
post #91

Calorie tracking is also about educating yourself about how many calories certain things are, so you can make better decisions. Like, oil is insanely caloric and can accidentally add hundreds of calories, but it's nearly impossible to eat too many greens. Once you learn this, then the tracking is just to keep you honest - your brain knows what to do but it lies to you when it wants to bend the rules and those little…

This was the huge benefit of weight watchers. I saw hundreds of people utterly shocked they could eat a literal mountain of salad for the same number of “points” (calories/50) as a tiny piece of something sweet or fatty.

They genuinely had no idea, and it changed their lives.

Re: I used AI-powered calorie counting apps, and they were even worse than expected

#200
post #128
post #107

Earlier quoted context omitted.

> most people underestimate the calories in alcohol Funny, I feel the other way around. I kept hearing about how much calories there are in alcohol, and then when I started calorie counting I didn't find it so high. Like 6 shots of gin are ~550kcal and enough to get anyone pretty drunk. Unless one is a regular heavy drinker it's not that hard to once in a while budget calories during the day to be able to get a few s…

The idea that a night out with 4-5 pints might clear my daily total requirement for calories was not something I had really considered until my late twenties.

In your early 20s your metabolism is that much higher as well that it can cope.
Post reply on HN