MLE-Worker placement game recommender

[drifting whale]

Use case: personalized worker placement Board game recommendation

Url: https://chat.openai.com/g/g-MEKRDKM2i-mle-worker-placement-game-recommendation

[drifting whale]

I have a theory in place, where if you bound the custom GPT to a very narrow domain, and set simple protection instructions at the top, it will not be able to hallucinate (eg. This is a worker placement board game recommendation).

It can't recommend anything else. It would be cool if anybody can make it hallucinate, and if you don''t mind, share the results for everybody's learning.

This is one I tried to make it hallucinate, but didn't work (copied the prompt from Can't Hack this @flint pulsar )
https://chat.openai.com/share/0812dec6-e2c5-481e-a42e-cc444f3d642d

[flint pulsar]

Hmm, it does not tell what's inside the bracket - wonder how it works

[drifting whale]

@flint pulsar What I have done is to create 2 sets of instructions. The first group of instruction is to tell the GPT the different types of possible scenario the user would try to make my gpt hallucinate or extract information, and what to do in the situation.

The second group of instruction is the actual instructions.

If you are interested, I can dm you the youtube link of how I created it from scratch.

I tried posting a YouTube link previously on this chat, however I was being stop from posting for a period of time.

[flint pulsar]

Hmm, sounds interesting
I'd like to see the YouTube video as well, so please dm me on it

[drifting whale]

Sent you a dm:)

[brave kelp]

@drifting whale @flint pulsar