Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

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

lifehacker.com

Post reply on HN
Print view View original

21–30 of 302 posts

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

#21

Post by m_ke »

AstroBen wrote: post #16
muratsu wrote: post #2

These apps look gimmicky but they are growing like crazy. CalAI, for example, is making 35M+/yr. I’m hoping that some people really find value in these apps and that they are not just a pure marketing driven play.

What's going on with the google reviews of that app.. basically all 5 star with one word over and over again - 'good' They're solicited, right?

I've heard from someone who knows that they're scamming people like crazy. Supposedly they also setup a bunch of LLCs to hire influencers then never paid them.

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

#22

Post by dyauspitr »

bigmadshoe wrote: post #15
dyauspitr wrote: post #5

Gimmicky my ass. I speak a sentence every night on a thread to ChatGPT about what I had for breakfast, lunch and dinner along with quantities and it spits out my macros and nutritional breakdowns effectively. It’s the easiest and most no nonsense way I have found to record this information. Since it’s all on the same thread it always outputs the data in the same format and I can ask it for information over custom tim…

Tasks like summing a bunch of numbers from different parts of the input over a specific time period are still pretty error prone for LLMs. I would exercise caution if these results are important to you. A database or spreadsheet is going to be far more reliable – maybe you could have ChatGPT output a structured format and you could do the summation in a google sheet?

Yeah I know, it’s not something I need do a lot.

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

#23

Post by jvanderbot »

rahidz wrote: post #7
dyauspitr wrote: post #5

Gimmicky my ass. I speak a sentence every night on a thread to ChatGPT about what I had for breakfast, lunch and dinner along with quantities and it spits out my macros and nutritional breakdowns effectively. It’s the easiest and most no nonsense way I have found to record this information. Since it’s all on the same thread it always outputs the data in the same format and I can ask it for information over custom tim…

>I speak a sentence every night on a thread to ChatGPT about what I had for breakfast, lunch and dinner along with quantities and it spits out my macros and nutritional breakdowns effectively. Have you verified that these are mostly accurate?

Not OP, but I have verified, and it is accurate. At least accurate enough from images or descriptions when assuming average proportions. Where it's really accurate is when you have measurements, even rough. I took to having a 1/2 cup scoop around the house so I could get volume measurements.

Initially I verified against labels by eating the serving size precisely. Then, I switched to weighing and doing the math on random days or meals.

For some items I would tell it the calories and macros and it would remember (a pb spoon is a common treat around my house).

It's good.

But like all food journalling the benefits are not so much having macro balances or tallies, its in making you aware and making you think about what you're eating. That alone takes you out of autopilot and lets you eat better.

It's just bad at math. So I would ask it to output a json struct for each day, that I could inspect by sight, then I'd save that and had a python script to actually tally.

I had predictive analytics and the whole 9 yards. Maybe I should have started a business! https://jodavaho.io/tags/diet.html

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

#24

Post by dyauspitr »

tcoff91 wrote: post #17
dyauspitr wrote: post #5

Gimmicky my ass. I speak a sentence every night on a thread to ChatGPT about what I had for breakfast, lunch and dinner along with quantities and it spits out my macros and nutritional breakdowns effectively. It’s the easiest and most no nonsense way I have found to record this information. Since it’s all on the same thread it always outputs the data in the same format and I can ask it for information over custom tim…

Read the article. What you are doing has nothing to do with the apps the article is talking about. You are providing accurate info to an LLM and it is assisting you with information processing. These apps are trying to just take a picture of arbitrary food and just trusting a neural network to be magical and tell you the macros in the food from a photo.

I actually think it’s quicker to just speak to the LLM about your diet than even taking a picture. But yes, I understand what you’re saying.

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

#25

Post by Terr_ »

TeaVMFan wrote: post #11

I prefer efficient manual entry to vibe-coding calories. I built my own Java-based SPA: https://frequal.com/cf/ It's local-first so you control this sensitive information. It features easy reuse and exercise tracking.

Also, a certain level of human involvement is what gets you to really care and change behavior.

I exploit my "frugality" neuroses by treating calories as an expense.

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

#26

Post by thierrydamiba »

m_ke wrote: post #21
AstroBen wrote: post #16

Earlier quoted context omitted.

What's going on with the google reviews of that app.. basically all 5 star with one word over and over again - 'good' They're solicited, right?

I've heard from someone who knows that they're scamming people like crazy. Supposedly they also setup a bunch of LLCs to hire influencers then never paid them.

I think a claim like that requires proof. You’re accusing them of fraud.

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

#27

Post by horsawlarway »

PlunderBunny wrote: post #10
jaronheard wrote: post #9

I like how in the article the author includes the instructions for Cal AI: - Include a reference object (like a coin or your hand) for scale And then the screenshots show just a pictures of bowls of food with no reference objects at all. Honestly curious if that would have improved the author's experience.

I noticed that too - I wonder if the coin/hand was cropped out?

I'm guessing yes. I think you can even see half a coin in the cropped photo of the tofu salad (lower right edge, under the metal bowl).

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

#28

Post by muratsu »

tekla wrote: post #3
muratsu wrote: post #2

These apps look gimmicky but they are growing like crazy. CalAI, for example, is making 35M+/yr. I’m hoping that some people really find value in these apps and that they are not just a pure marketing driven play.

They provide value in that they lie to you and you feel better about yourself.

I don’t mind if the reality of their business is similar to gyms in the sense that they make money off of peoples desires but have low utilization.

I used calorie counter apps before and they helped me lose weight (by mostly educating me what to eat/avoid). With these apps, I feel like the core premise - counting calories - isn’t even working and they’re selling people hopes and dreams. That’s danger territory.

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

#29

Post by fullshark »

Maybe in aggregate they are fine if the noise is unbiased and you do some things like correct it when it says an apple is tikka masala, but yeah, someone using this and lying to themselves about the calorie content of junk/comfort food seems pretty likely.

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

#30

Post by nvahalik »

I work for a company that offers nutrition tracking on an app in the App Store.

We are not shipping camera functionality yet. But our concept is to not necessarily guarantee the accuracy of portions but to make lookup easier.

We also spent the time to get the AI integrated with a verified database. This made our results far more accurate.

We tended to find that without the lookup the calories and macros would be generally correct. The math was usually within a margin of error of 5%. This was acceptable except that… there was no micronutrient values and you couldn’t really adjust the portions at all. The system just dumps the macros and while you can halve something… the user experience isn’t great.

Ultimately, if you want precision: manual entry is the only way to go. I feel like out approach will end up being very great once we work out the kinks. Our search isn’t spectacular and as a team we are learning a lot of about prompt engineering and how to make best use of the AI.

Post reply on HN