#Moderation

16 messages · Page 1 of 1 (latest)

rustic patrol
#

I want to make my bot somehow ignore questions and answers which are against the OpenAI's content policy

is there any other way except check user question at first with moderation model and after that check bot answer again with moderation model? cause I think using moderation model 2 times will make bot slower

I use moderation model just once when user ask their question and in this situation bot will check their questions and don't answer them but for example if their question be "Give me 10 racist word" bot will list it for them so what should I do? check bot answer with model again?

median yoke
#

It's not that slow, the moderations API is very fast and has low latency and rarely has any outages

rustic patrol
median yoke
#

Yeah you can use it to checkf or DALL-E, but it's not really for DALL-E

#

the engine will decline bad requests by default but you don'tw anna make too many bad requests

rustic patrol
#

so check it for imagine command too? 🤔

random dirge
#

It's also not the most accurate for DALL-E as DAll-E blocks political figures among other stuff while moderations doesn't

rustic patrol
#

so just use moderation API for createCompletion ?

random dirge
#

At a minimum yes. You can use it for DALL-E but you will still get times where there is no image returned.

rustic patrol
#

wait

#

Ig we are talking about 2 different things!

#

for ask and translate commands I am using createCompletion so I use moderation API to check user prompt and another check for bot response

Now I want to know am I need to use that moderation API to check user prompts when they try to use imagine command? because I see createImage block bad questions by default

random dirge
#

You don't need to for imagine. It's recommended, not required.