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

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

amazon.com

Post reply on HN
Print view View original

61–70 of 768 posts

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

#61

Post by ryandamm »

Retr0id wrote: post #55
AlecSchueler wrote: post #35

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.

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.

Why not have a separate chat request to apology-check the responses?

Not my original idea, there was a link from HN where the dev did just that.

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

#62

Post by boxedadin »

Retr0id wrote: post #55
AlecSchueler wrote: post #35

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.

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.

You could go full circle and ask OpenAI to determine if another instance of OpenAI was apologetic.

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

#63

Post by AnarchismIsCool »

Retr0id wrote: post #55
AlecSchueler wrote: post #35

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.

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.

Time to create on algorithm that operates on the safety flag boolean to optimize phrases to bypass it

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

#64

Post by rany_ »

AlecSchueler wrote: post #35

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.

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".

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

#65

Post by Retr0id »

ryandamm wrote: post #61
Retr0id wrote: post #55

Earlier quoted context omitted.

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.

Why not have a separate chat request to apology-check the responses? Not my original idea, there was a link from HN where the dev did just that.

Sounds like a great way to double your API bills, and maybe that's worth it, but it seems pretty heavy-handed to me (and equally not 100% watertight).

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

#66

Post by csours »

evan_ wrote: post #6

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 li…

Brown

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

#67

Post by ganeshkrishnan »

evan_ wrote: post #6

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 li…

You can see in this product: https://www.amazon.com/analyze-generate-product-avoiding-tra...

ChatGPT is refusing to generate titles with trademarked names. So most likely they are prompting something like "competitor product: rephrase the title"

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

#68

Post by titzer »

evan_ wrote: post #6

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 li…

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?

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

#69

Post by nxobject »

lancesells wrote: post #32
siva7 wrote: post #21

Earlier quoted context omitted.

If we don't get soon effective measures to separate AI bots from humans this will be the end of the Internet as we know it.

Amazon brought this on themselves by allowing all of this garbage in the first place.

I'm genuinely shocked that there's no immediate disincentive to all of these shell vendors, other than pitting search results with varying levels of sponsorship against each other.

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

#70

Post by em-bee »

csours wrote: post #60
sandworm101 wrote: post #37

Earlier quoted context omitted.

>> unattended LLM do anything for your company at all. It can, and will, fuck up dramatically sooner or later. So, just like any other random employee?

The employee generally knows they fucked up and can escalate the issue. Discussion on whether or not this actually happens will follow in comments below.

or if they don't know at the time, they may eventually realize it later and react accordingly.
Post reply on HN