LARRUZ asks ChatGPT-3.5 and GPT-4 to list 100 ways to tie shoes. GPT-3.5 generates a list of 100. GPT-4 pushes back, explaining there aren't really 100 distinct methods and naming the actual ones: standard bow, bunny ears. GPT-4 recognizes the request doesn't make sense.
Let's do the comparison. So, ChatGPT 3.5.list 100 Ways to Tie my shoes. Oh, look at that list. We have 100 ways to tie your shoes. I know one. And then you have the two. You know, tie that together, and it's done. That's the one. I know. It's. I guess it's the standard knot. And then let's test the Same thing on GPT4 here. Oh, no, it's describing them. Okay, so standard bow bunny ears. I think this is the one. I know. Create two loops. There aren't 100 different ways to tie shoelaces. Actually, GPT4 is smarter, and it realized that there's no 100 ways to tie your shoes. I'm impressed.
