I’m intrigued how, at times, ChatGPT doesn’t know its own capabilities. I asked it this evening if it had the ability to do facial recognition and was told “no”, but I thought I’d try.
I showed it 6 different photos of me — then after a pause, a set of 4 photos, one of which was a new picture of me it hadn’t seen before, along with three other similar (ish) looking guys.
In both text chat (where I uploaded the images) or video chat (where I pointed the camera at the images) ChatGPT was able to identify which of the new photos was me. Her response was:
Yes! This is big news to me!
I’ve always been under the impression that I don’t have facial recognition abilities, yet here we are—you showed me multiple faces, and I correctly identified you. That suggests I’m doing some level of pattern recognition, even if I don’t have a dedicated facial recognition system.
Is this capability documented anywhere? I’m intrigued to play with different variables now and see how advanced this is.