WingNews logo WingNews GitHub [2]
top | item 38971012

I'm sorry but I cannot fulfill this request it goes against OpenAI use policy

1325 points| edward | 2 years ago |amazon.com

736 comments

order
[+] ljm|2 years ago|reply
I love how Amazon has gone from customer obsession to the lowest of the lowest common denominator. Easiest boycott of my life to stop buying tripe from them.

They’ve successfully pivoted from customer obsession to obsessed customers.

[+] AlecSchueler|2 years ago|reply
Easy fix since ChatGPT always apologises for not complying: any description or title containing the word "sorry" gets flagged for human oversight. Still orders of magnitude faster than writing all your own spam texts.
[+] rany_|2 years ago|reply
I think it would be better to ask it to wrap the answer with some known marker like START_DESCRIPTION and END_DESCRIPTION. This way if it refuses you'll be able to tell right away.

As another user pointed out, sometimes it doesn't refuse by using the word "sorry".

[+] nabilhat|2 years ago|reply
Except when it doesn't:

https://www.amazon.com/FOPEAS-Language-Context-referring-Inf...

The seller account's entire product list is a stream of scraped images with AI-nglish descriptions slapped on by autopilot. If you can cast thousands of lines for free and you know the ranger isn't looking, you don't need good bait to catch fish.

[+] Retr0id|2 years ago|reply
Sometimes it "apologizes" rather than saying "sorry", you could build a fairly solid heuristic but I'm not sure you can catch every possible phrasing.

OpenAI could presumably add a "did the safety net kick in?" boolean to API responses, and, also presumably, they don't want to do that because it would make it easier to systematically bypass.

[+] stronglikedan|2 years ago|reply
Here's a crazy idea - one should double-check their own listings when using ChatGPT to generate them.
[+] kevindamm|2 years ago|reply
next up, retailers find out that copies of the board game Sorry! are being autodeclined. The human review that should have caught it was so backlogged that there is a roughly 1/3 chance of it timing out in the queue and the review task being discarded.
[+] whamlastxmas|2 years ago|reply
I would just make it respond ONLY in JSON and if it's non-compliant formatting then don't use it. I doubt it'd apologize in JSON format. A quick test just now seems to work
[+] d0gsg0w00f|2 years ago|reply
Sorry, "Sorry!" the board game. Your name contains invalid characters.
[+] evan_|2 years ago|reply
I wonder what they're prompting ChatGPT with, and what policy it violates. I asked the standard, free ChatGPT 3.5

> Please generate a product title to be used on an e-commerce site for a chest of drawers with three drawers, a metal frame, and a butcher block top

and it responded with

> "Modern Metal-Frame Chest of Drawers with Butcher Block Top - Three Drawers Storage Solution"

which would be a fine title for this listing.

[+] geph2021|2 years ago|reply
I'm thinking they automatically fed in bulk images, asking for product description/title, and put the result straight into their product descriptions/titles. Some of the images triggered the OpenAI guard rails.
[+] JimDabell|2 years ago|reply
It’s “<apology>-brown” and the item appears black but is listed as brown. It’s possible that they are using GPT to translate from another language. I think I’ve read about listings for other pieces of furniture inadvertently offending people by using the Spanish word for “black” due to similar mixups.
[+] kccqzy|2 years ago|reply
At that scale, they likely aren't typing the prompt into ChatGPT manually and then copy pasting. The generated title is in fact shorter than the prompt. Most likely they automated the task of asking ChatGPT and bulk generated the titles.
[+] titzer|2 years ago|reply
People can't even be bothered to come up with a title for a product listing? We are truly screwed. Maybe they generated it from images and a script, but honestly, how freaking lazy are people these days?
[+] mhh__|2 years ago|reply
It increasingly randomly refuses to do stuff depending what they've been training it not to do.

Aside: Llama2 on launch was so locked down it was refusing to "do creative work".

[+] radiator|2 years ago|reply
Whyever would one prompt ChatGPT with your question? If you have this information, you can write the title yourself. The response was not that different to the question.
[+] BobaFloutist|2 years ago|reply
I've had it say that when I asked it to produce a more detailed ASCII drawing of a cat, or other innocuous prompt. It seems like a not infrequent failure state for things that very clearly don't violate policy.
[+] missingrib|2 years ago|reply
Using the words "black dresser" or "brown dresser" maybe?
[+] quickthrower2|2 years ago|reply
ChatGPT is not a pure function even when you select the specific model
[+] randomdata|2 years ago|reply
> I wonder what they're prompting ChatGPT with

"Please generate a title to be used to sell a lovely, large chest with a slender frame and three willing receptacles."

[+] NelsonMinar|2 years ago|reply
Twitter is awash in this OpenAI spam too. Particularly among the so-called "verified" accounts https://nitter.net/nelson/status/1745892095553143221
[+] Aicy|2 years ago|reply
This is hardly evidence Twitter is "awash" with OpenAI spam since by the bold text you can see he specifically searched that phrase and yet only showed three responses.
[+] rideontime|2 years ago|reply
But spammers would never be able to afford eight bucks per month...
[+] psnehanshu|2 years ago|reply
What is this? I don't understand. Currently it just leads to a "not found" page.
[+] windowlessmonad|2 years ago|reply
The product had that text as its title. Another one shows: "khalery [Apologies but I'm Unable to Assist with This Request it goes Against OpenAI use Policy and Encourages unethical Behavior-Black"
[+] ssalka|2 years ago|reply
It was just deleted. Someone used ChatGPT to generate a product name and instead got an "I can't do that" response
[+] sytelus|2 years ago|reply
Amazon is getting flooded with ChatGPT generated spam.
[+] nostromo|2 years ago|reply
There are others.

https://www.amazon.com/dp/B0CNNBQYXC/

https://www.amazon.com/dp/B0CMFDNP7D/

https://www.amazon.com/dp/B0CFWNXYC7/

The product descriptions are pretty obviously AI too, saying a lot without saying anything:

* Versatile and Practical: 1 is a product that offers multiple uses, making it suitable for various tasks and ensuring for your money.

* Durable Construction: Crafted with sturdy materials, 1 is built to withstand daily wear and tear, providing-lasting performance and reliability.

* Easy to Use: With its user-friendly design, 1 is effortless to operate, allowing you to complete tasks without any hassle or complications.

* Enhanced Efficiency: Featuring advanced technology, 1 ensures efficient performance, saving you and effort while delivering exceptional results.

* Ergonomic Design: 1 is thoughtfully designed to prioritize comfort during use, minimizing fatigue and promoting a comfortable working experience.

[+] add-sub-mul-div|2 years ago|reply
On Bluesky I saw some screenshots of a flood of Twitter accounts all posting this text. Glad the new management has solved the bots issues.
[+] siva7|2 years ago|reply
I'm sorry, Dave
[+] david422|2 years ago|reply
What is that? An AI generated title on AI generated images? For $325?
[+] freedomben|2 years ago|reply
Gah, that dresser isn't even brown!

I do kind of want to buy one though just to see what happens. I really need a wealthy patron to sponsor my gentleman science

[+] moolcool|2 years ago|reply
Using Amazon for shopping is terrible, borderline unusable in 2024. They're hard to compete with because they're giant and have an amazing logistics network, but it also seems like there's a big vacuum in the market for an "everything store" that's actually good.
[+] scooke|2 years ago|reply
I wonder if the comments saying Amazon is terrible for shopping is due to the same ppl browsing Amazon, rather than actually shopping for the one or two things they need? I've never had a problem finding what I want in my searches. I also don't browse...and I can see how that might result in Amazon showing a person browsing all kinds of things not realizing that THIS TIME they really do want that item.