It is worth noting that GPT-4 can accept a prompt of text and images, which—parallel to the text-only setting—lets the user specify any vision or language task.