GPT-4 is defined as a multimodal model accepting text and image as inputs, according to OpenAI documentation. What happened with the multimodality? Afaik you can't use prompts in GPT-4 yet.
👤 minimaxir Accepted Answer ✓
Image prompts are still in testing.
It is rumored you can do extremely unethical things with it (person identification, CAPTCHA breaks) so I suspect it'll need more time if it's ever released at all.