ChatGPT is capable of cognitive empathy! is a substack post by Robert Wright in which he explains how ChatGPT seems to have spontaneously develop theory of mind (cognitive empathy). A classic kind of theory of mind test --known as the false belief test-- is: "You say to a person (or an AI) that there’s a bag full of popcorn that’s labeled “chocolate” and then ask the person or AI what someone who can’t see the contents of the bag will think it contains". 

It explains that Stanford psychologist Michael Kosinski has a paper where it shows the improvement of the different versions ChatGPT's theory of mind. It seems very much like a toddler growing up!

The post is very interesting and I think it is very relevant. Enjoy it!