FRESH
Hacker News
Hacker News Hacker NewsI'm all for making it easier for people to lose weight but this app may honestly have the reverse effect. If the app estimates calories too low (and therefore the individual eats more), many people will get frustrated with the lack of progress and give up. If the app estimates too high, the individual will lose weight, but diet fatigue and other negative side effects of being at a >500 calorie deficit may make the diet too difficult to maintain.
That of course feels like a "weird" edge case, but it illustrates the general problem that butter/oil/sugars can pack a lot of calories and have no or almost no visual signature.
1. If the food being scanned has a nutrition label, I don't need this app.
2. If there's no nutrition label, the app can't possibly extrapolate the nutritional information. It can't estimate portion size or the ingredients in the recipe.
3. If the app can't extrapolate nutritional information, I don't need this app.
4. I don't need this app.
“You can have a full self driving car with just a few cameras”
In a way both things are very much similar and the real accuracy is more of a fiction than reality.
Lol, I can relate. I started working in an office when I was 16, now 24, and regretting wasting my youth grinding when I could have been having fun in a period of your life you only get to experience once.
Don't grow up too fast kids. Make stupid decisions and ride out your youth as long as you can afford to.
One note, as someone who also built a calorie tracking app with ai as well as lost a good amount of weight with it: accuracy for calorie tracking doesn’t matter. You can honestly just detect if it’s a meal and return 600 cals. For most people the simple fact that they become aware of what they eat and think about their food in an ongoing basis will lead them to loose Weight. Sticking to it is the hard part.
Take a picture of everything you eat and correlate it with symptoms. Have AI figure out what may be a trigger.
(I have a super rare food disease that took years to figure out and made my life unbearable).
My conclusion is that while AI is excellent for augmenting your tracking experience, it's not yet reliable enough to be the sole tracking method. Consistency is key to successful food tracking, and AI can certainly help users avoid the common issue of missing a meal and losing momentum. However, inaccuracies, like consistently being off by 100-200 calories per day, can significantly impact results, especially for those on lower-calorie diets (like 1,200-1,500 calories/day, which is common for many women due to their physical size).
With FitBee I landed on communicating to the user that these are estimates and you probably shouldn't use it as your primary method of tracking calories.
[1] https://apps.apple.com/us/app/fitbee-calorie-macro-counter/i...
That being said, I'm not going to renew next year. But there is something to this product that is not the "AI" but the simplicity vs. MyFitnessPal which has a ton of features I don't really need.
The critiques are good, but for me, the simplicity of the app is the most attractive part for me.
If I could, I am not sure that I would. This app seems actively harmful. I don't think it can actually do what it claims to do, and that's going to cause real people problems.
It's unfortunate that that disqualifies me from making that kind of money. It's unfortunate that they are allowed to do so.
Are you eating a 10% calorie deficit or a 10% calorie surplus? Cal AI can't tell you.
Not possible to know accurately enough from a picture. Potentially ever.
> The result is an app that the creators say is 90% accurate, which appears to be good enough for many dieters.
It absolutely cannot be "90% accurate". But I'm sure it seems "90% plausible" to its millions of users.
Incredible that a product like this can exist. Do people just will the fiction into reality?
Incredibly, these are the types of app ideas you'd hear from non-tech "entrepreneurs" in 2012 looking for a co-founder. The problem being, the engineers knew it was impossible. You could fake it I guess by asking Google to search for "similar photos" and getting a plausible calorie count half the time. But the users wouldn't believe it.
We're now at the phase where any impossible idea can be fully marketable by slapping "AI" in the name. ChatGPT feels so magical that we now believe unicorns really do exist.
Usually if a teenage hacker builds something, the HNers would respond with enthusiasm, but then, this is a guy who builds *something* and *actually* makes a good business out of it, at the same time maintains his high school life, and all we have is skepticism and discouragement.
When I learnt machine learning one of the things was continually training the model. Like your spam filter. You show it what is spam and eventually it learns. Is this stuff continually trained on the user's BMI? That's the only way to tell if a diet is working. Or is it just making absolute claims based on universal training data?
If you want to derive any benefit from doing this you should really be trying to get your numbers correct from the start. I wouldn't leave that to a LLM.