We are seeing a very high level of invalid_request_error returned in the past 12 hours for the DALLE2 API. The error looks like this:
{
"error": {
"code": null,
"message": "Your request was rejected as a result of our safety system. Your prompt may contain text that is not allowed by our safety system.",
"param": null,
"type": "invalid_request_error"
}
}
We screen every call via the moderation endpoint and that comes back negative. We didn’t have this until recently.
Your call may be rejected based on the prompt or the resulting generation.
Also based on my usage of DALL-E 2, it doesn’t use the moderation endpoint to filter out prompts.
Prompts and images are filtered based on our content policy, returning an error when a prompt or image is flagged. If you have any feedback on false positives or related issues, please contact us through our help center.
I sure he will be very happy to hear that his name is unacceptable. Seems like a slightly crude approach to moderation for such an advanced tech company. Anyway thanks very much for helping to cast some light on this. Appreciate it.
That is obviously an unnecessarily aggressive OpenAI Dalle-E filter.
Note: Changed the topic title, however, based on how very busy the OpenAI staff are and that they are rarely here, it’s doubtful they will read this message anytime soon.
Thanks for confirming and agree about the title. Hope someone from the OpenAI team sees it and in the mean time people that have tmesis names will just have to be sad that Dall-E won’t draw images of them.