[mewpmewp2]

I mean more interactive way, not notifications being sent.

E.g. you start interaction by asking the fridge - I am a bit tired, what is an easy thing to cook based on what I have in the fridge and I am stopping by the store, is there anything I should buy to have a few more options.

So GPT vision takes image of what is in the fridge and then tries to solve for that question. If user is allergic to something that would be in GPT's prompt.

[arter4]

Ok, but if it's interactive I can open a fridge, take a picture of it and upload it to ChatGPT. You wouldn't need a smart fridge for that.

You would need a smart fridge if you wanted to directly connect the fridge with ChatGPT. Which is what I assumed in the previous comment.

[mewpmewp2]

You may be at work, and store is on the way to work. Also what you described is quite many steps, and the images have to be taken at certain angles to capture everything.

It would be directly connected to GPT and you can either use it as mobile app or with other devices or fridge itself directly.

So for example you are at work, about to leave home, store is on the way and then you ask this question from the app. Fridge takes photo, forwards it with certain prompts to OpenAI APIs and then gives you the response.

[arter4]

So you're saying a fridge takes a photo (of itself, essentially... a selfie?), forwards it with certain prompts to a GPT and then gives you the response. But as I said earlier I might not like certain foods, or I may actively avoid certain foods for medical or religious reasons.

Now, if this was an explicit user-GPT interaction, you could prompt the GPT with "hey, I'm Muslim, what can I cook with this stuff?" or "hey, I'm lactose intolerant, what else could I buy?". You do need to trust the OpenAI provider, but not the fridge.

Instead, the fridge automatically talks to GPT. So if you want to avoid "dumb" suggestions you wouldn't be able to follow, you would have to tell your fridge you're Jew or Muslim or lactose intolerant or diabetic. Don't you see an issue with this?

I'm not "tinfoil hat" paranoid, I don't think people are out there to get me. But a certain dose of skepticism when it comes to data usage should be healthy, especially when many companies have been known to mess up this aspect (from Cambridge Analytica to Roomba employers sending pictures of a woman on the toilet to Facebook, and much more). You cannot avoid this 100% unless you go to extreme lengths, but if you can avoid sending data to one more company, why not?

[mewpmewp2]

In fridge it takes images from different angles, and different floors to make sure it is possible to identify all the items that exist in that fridge.

If you don't like certain foods, you can just customise the prompt that fridge sends to OpenAI. You can customise it through a web app or mobile app.

> you would have to tell your fridge you're Jew or Muslim or lactose intolerant or diabetic. Don't you see an issue with this?

What's the problem with storing your food preferences somewhere? People use dieting apps all the time on mobile.

> You cannot avoid this 100% unless you go to extreme lengths, but if you can avoid sending data to one more company, why not?

You are already sending your data in thousands of different ways and exposing yourself at any moment. I'm not saying you should do more of it, but food preferences seems like a very minor drop in the bucket of all that you are already exposing.

Most people store their images in a cloud, which already indicate amazing amount about them. Just using a smartphone is 1000x worse than having food preferences stored somewhere.

[arter4]

Sure but it's something more for barely any actual advantage.

I accept that when I Google something Google and maybe other companies know what I searched, because Googling things is useful. I cannot bring an encyclopaedia and a detailed map of the whole world with me all the time, so it's a tradeoff I can accept.

I accept that, on the rare occasions I turn geolocation, someone will know where I am, because if I do turn it on, it's probably because I got lost. Then I turn it off and off we go.

Divulging food preferences for... an AI that suggests me what I should buy doesn't look like it's worth it. And it's not just food preferences. Remember people store medications in their fridge. A fridge so advanced it can recognize any food is probably capable of reading labels and knowing I bought medication for X,Y and Z. Which is definitely something I'm not explicitly telling people (or things) without a reason.

And don't forget the power of correlating data. Correlating food preferences with other data you could easily understand if someone is ill, pregnant, whether they follow a certain religion (think Ramadan, or not eating meat on Friday,...), whether they are living with someone else (buying twice the amount of stuff they usually buy?) and so on.

This is all stuff I try to avoid telling everyone. Yes, my doctor could be hacked and people could know I'm on medication anyway (hypothetically), but why divulge that voluntarily for essentially no reason?

[mewpmewp2]

Honestly you are just describing to me more exciting features. The correlations with potential illnesses, etc. Have all of your data piped into single location to use AI to improve your health.