Hacker News Reader

Comment by rolosa | Hacker News Reader

I would not trust the AI to order food off of these websites for me because the labeled data itself is pretty much shit. Lots of dishes are unlabeled, have really bad pictures or no pictures at all, and the worse is the ones that have AI generated descriptions which are often wrong or not helpful. I'm totally fine with letting it just use my saved credit cards and delivery address (I assume the AI doesn't see them) but I can't trust it to order the food I actually want.

E: As an example (n = 1 etc): I am vegetarian. I have never ordered meat dishes, ever. Yet all of these food delivery apps suggest meat dishes for me.

View on HN

I have never ordered meat dishes, ever. Yet all of these food delivery apps suggest meat dishes for me

This has been my experience too, with Uber Eats. It recommends meat dishes, every. single. time even though I have never ordered a meat dish, even once. This is the absolute basic, minimum help a food recommendation algo should do, and it still doesn't. I wonder why. Maybe vegetarians are too small a percentage of customers for these apps to care?

> I assume the AI doesn't see them

Why?

doordash is regularly audited for PCI, they almost certainly store raw CC#s and follow modern tokenization standards. The best the agent would see is the name you stored for the card if you have multiple cards it can pick from in the account.

You can pull the PCI compliance documentation from their trust center if you are a partner https://trust.doordash.com/

For addresses, it's a almost certain the agent would see those since those are visible in the UI when ordering and required for selecting a destination for the order

It's plausible, but also, AI systems have a long history of being plugged in on the wrong side of security boundaries.

Remember when you were able to take over any Instagram account by asking the helpful support AI to send a password reset to an email address of your choice?

There’s a difference between their AI and your AI. For better and worse.

>Remember when you were able to take over any Instagram account by asking the helpful support AI to send a password reset to an email address of your choice?

Wow, first I've heard of that. Damages must have been immense for that level of incompetence occasioning financial harm. Or do those laws only apply if you can't hire very expensive lawyers to blame a computer?

It's better they don't because considering how Claude reacts when you give it a meaningless SSH login password for a local box, the moment it learns your CC# and CVV2 it will probably call your bank and cancel the card instead of ordering any food.

are you saying Claude is not a strict constructionist when it comes to Constitution AI? :)

> AI, I asked you to check the docs for the latest version of the API, yet you didn't. Why?

> You are completely right! That was my miss, I have no excuse. I've now looked at the docs and implemented it.

Now imagine how triggered you'd be when the AI decides to get you a vegan burger because the chicken wrap is on the second page and it know you're a health conscious person.

Reply on news.ycombinator.com