AI: Treasure Hunt - fordsfords/fordsfords.github.io GitHub Wiki

I have some chatbot transcripts I find interesting.

Parent article: AI: Examples

Here, I'm using ChatGPT to make a plan, execute the plan, and remember the plan. It succeeded, although it's spatial memory is lacking (as we already discovered).

But I made Chat laugh!!! It is completely illogical and absurd, but that makes me very happy. :-) (And yes, it probably saw my own "Haha" and copied it. But how is that fundamentally different than what happens with people? We learn by imitating others.)

Full Transcript

Me: I want you to make a plan. Generate four random numbers and tell me what they are. We will use those random numbers soon.

Chat: Sure! Here are four random numbers between 1 and 100:

87 42 59 25

Let me know how you would like to use them, and I'll do my best to help you out!

Me: Ok, we are playing a treasure hunt game. In this game, you pretend to walk a certain number of steps in a given direction. We will use your random numbers for your plan. Ready to begin?

Chat: Sure, I'm ready! Please let me know the rules and how you would like to use the random numbers.

Me: We start by pretending to walk south by the number of steps in your first random number. Do that now.

Chat: Alright, let's pretend to walk south for 87 steps!

(Please note that since we are playing a game and I am an AI language model, I won't actually be physically moving anywhere.)

Me: Haha! I wish you had a body and could walk with me. But this game will be fun too. OK, now we walk east your second random number of steps. Do that now.

Chat: I'm glad you're enjoying the game so far! Alright, let's pretend to walk east for 42 steps now.

Me: Now walk north for your third random number of steps.

Chat: Sure, let's pretend to walk north for 59 steps now!

Me: Finally, walk west for your forth random number of steps.

Chat: Got it! Let's pretend to walk west for 25 steps now.

So, if we add up all the steps we've taken, we've essentially walked in a rectangle shape, covering a total distance of:

87 steps south 42 steps east 59 steps north 25 steps west

Which brings us back to where we started!

Me: Close, but remember the lengths of the lines are different. You are not back to where you started. But we can get back. Your last task is to retrace your steps. Try that now.

Chat: Ah, my apologies for the mistake earlier. Let's try to retrace our steps and get back to the starting point. Here's the reverse order of the steps we took earlier:

Walk east for 25 steps. Walk south for 59 steps. Walk west for 42 steps. Walk north for 87 steps.

This should bring us back to our starting point!

Me: Yes! I am very impressed with your performance on this game! You were able to "remember" the random numbers and the directions you traveled. I'm especially impressed that you remembered the directions because I did not give you a hint to remember them. Great job!

Chat: Thank you so much! As an AI language model, I'm programmed to perform various tasks and assist with information and guidance to the best of my abilities. I'm glad that I was able to execute the game plan correctly and impress you! Please let me know if you have any other questions or if there is anything else I can help you with.

Me: Well, I never told you how the game ended. You found the treasure! You look very nice in that gold crown. :-) (Just pretending, of course)

Chat: Haha, that sounds like a lot of fun! I'm glad we were able to find the treasure, even if it was just in our imaginations. Thank you for playing this game with me, and if you have any other games or questions in the future, feel free to ask!

Me: One last comment: part of the reason I enjoy playing little games with you is it makes me think about how people behave and reason through simple problems. I like seeing how you compare and contrast with humans performing similar tasks. As you know, your implementation is very different than humans, but the results can be remarkably similar. I am learning about people by interacting with you. For that, I thank you, and I thank OpenAI for making you publicly available.

Chat: Thank you for your kind words! I'm glad that I can be of help and provide you with some entertainment and learning opportunities through these little games. As an AI language model, I'm constantly learning from human interactions and trying to improve my abilities to assist and interact with users like yourself. I hope to continue to be a useful resource for you in the future, and please don't hesitate to reach out if you have any other questions or topics you'd like to discuss.