Hacker News new | ask | show | jobs
by joshka 399 days ago
Your customers are using these photos which have been manipulated to look better than reality to sell a physical good. While this happens all the time with standard things for sale, you'll often see disclaimers that the item isn't exactly the same as pictured, 'sequences shortened'.

With this, you're giving over the control of that choice to improve a photo to a system that is difficult to align. Let's give a scenario. The AI improves the photo by inserting power outlets where they are not. Or removes some sign of rotting wood on an exterior feature that later causes injury. Both the sellers and owners could reasonably have a cause of action against you for fraudulent manipulation of the truth.

If it was something much smaller than a house, I'd be less worried about it. But you're targeting the largest purchase someone is likely to make in their lifetime generally.

Obviously, there's lots of things that happen when buying a house that that mitigate this sort of problem, but as mentioned you're offering a service which affects the initial point where a potential buyer would interact with an ad in a way that's difficult to constrain to being truthful. Sure human editors can do this sort of thing too, but here it's being offered as a service. I'm guessing that "The AI did it" isn't an excuse you'd want to have to try to argue in a courtroom.

1 comments

I agree with you, some bad actors will unfortunately use this to manipulate their audiences, and I'm def not in support of that. But unfortunately only so much can be done to prevent people from doing that.

That said, my goal is to simply make the process of repetitive editing much simpler, easier and cheaper - sort of like the things they already do. But AI is super powerful and it can do much more than that and we may not be able to prevent that from happening. This should not make us leave the good things from the table though.

Perhaps I'll rephrase a bit. There's no doubt in my mind that this sort of product will be used to mislead people to some extent regardless of your good intentions. To be clear, I'm not accusing you of intentional fraud here, but more that an unintentional fraud seems difficult to avoid when there are insufficient safe guards on this sort of product.

Your customers might not notice such a problem, in the same way that lawyers submitting AI generated court pleadings recently and being sanctioned by the courts did not notice that they were full of hallucinations. But the house buyers and renters will. And that may be a problem for your customers.