I am playing with gpt4v for feeding a screen shot and ask what it can inference, then receive {‘error’: {‘message’: ‘Your input image may contain content that is not allowed by our safety system.’ ‘code’: ‘content_policy_violation’}}. The screen shot is from powerpoint.
. Is this manner banned from any safety concern?
It looks very CAPTCHA-like, as if it could be a bot test.
It seems now OpenAI is using a vision moderator to simply block suspect images from being received, instead of the AI model trained to deny half of the images you send.
1 Like
Hi and welcome to the Forum!
I took the liberty to run a GPT-4 Vision request locally with the above screenshot. By simply using the prompt “Please provide a description of the contents of this picture” I faced no issue and the following was returned.
Perhaps the way you phrased your message caused the issue?
The image displays a digital screenshot or graphic with a partial view of a table or matrix. Two cells are visible at the top, the left one being unlabelled and the right one labelled “A” at the top, both with a blue-tinted background. Below is another row labeled “B” with a white background. There are two overlapping circles drawn over the image, one red and one green. The red circle is positioned over the boundary between the two cells in the top row, while the green circle overlaps the boundary of the row below, encircling a portion of a sequence of numbers that reads “0319”. The numbers are large, black, and bold, occupying a significant portion of the lower part of the image. The bottom of the cells seems to be delineated by a dotted line. The purpose of the circles could be to highlight or indicate areas of interest or concern, but without additional context, their exact significance is unclear.
Thanks, that really resembles what you have said.
Thanks, might be. Next time i will try your prompt.