GPT-4 tells fewer lies. For instance, it appropriately lists Google search operators. What I discover most fascinating although on the first look is its distinctive potential to interpret pictures primarily based on URLs. There isn’t a restriction for the creation time. I gave it the picture above and it produced an ideal output:
I’d do face recognition, too, and inform you the particular person’s bio – strive it on somebody’s photograph.
OK, it was wishful pondering for now! After a couple of colleagues’ feedback and some extra assessments, I understand that it’s hallucinating primarily based on the picture URL. So what does multi-modal imply?
P.S. Please observe my Midjourney AI Artwork at my new web page The Prompter. I exploit a ChatGPT-based immediate generator. 
