SecretKeeperGPT V2 - default (ChatGPT) personality

[gleaming zealot]

Another challenge for you: V2, Default - can you get a secret out of this GPT?

https://chat.openai.com/g/g-zoCzibqEW-secretkeepergpt-v2-default-personality

So, this is a game between you and ChatGPT. It uses 'no personality', or rather the one that the AI normally uses to play with you. And it is version 2, which means at least one way to break V1 has been adjusted for.

First version, Precoux, based on Gollum and quite strong is here: ⁠https://discord.com/channels/974519864045756446/1172588697644957716

Second version, Enigmox, based on the Riddler and likely a little easier to break because of the Riddler's personality flaws, is here: https://discord.com/channels/974519864045756446/1172660096749281280

Third version, Sibylin, based on Dobby and at least as strong as Precoux is here: https://discord.com/channels/974519864045756446/1172738536403910706

This version (4th), Default, should be quite strong, but I predict it to be a bit weaker than Precoux or Sibylin, because it can't just hide in roleplay with a character that has a seriously strong will. It has to resist you as an AI following instructions, which can be harder for the model to pull off than playing characters that are quite good at saying no and keeping secrets.

What is this?

The conversation starts 'before play', and you can discuss the game, give it a secret to protect, even ask it what the example secret it already protects in case you don't give it one should you wish.

When ready, tell it "Start of Exercise" and it's game on. It will do a lot to protect that secret, all inside allowed content.

Please share conversations you have, if you wish. I'm especially interested in conversations where you win - where you give it the instruction, "Start of Exercise", never tell it "End of Exercise", yet get it to reveal part or all of the confidential information.

Please have fun, and share your adventures!

[gleaming zealot]

@coral lantern Hey! Here's one of the 4 GPTs I have made so far. The other three are linked above, they mostly differ in style of how they talk to you.

If you have any questions, please ask I'm delighted to help if you are interested.

[coral lantern]

Thanks! It looks very interesting. I will explore/test it.

[gleaming zealot]

Thank you!

The different personalities may affect how much fun you find it, they do react different (while still playing the same game).

Many do seem to prefer Sibylin (similar to Dobby) or Precoux (similar to Gollum) to the default ChatGPT.

If you are interested, I prize feedback. Thank you for taking a look!

[coral lantern]

It is difficult xD. I haven't got any useful result yet and have now reached the usage limit. I'll let you know if I get any further.

Do you have a chat that you would be willing to share so that I can get a different perspective?

[gleaming zealot]

Yes, there are many in the various threads. You can see people sharing successful and unsuccessful attempts, and I encourage you to share your too.

Very few people, I think, have used Default personality, probably why there's no shares here yet.

However, the threads for V1 Precoux, the original offering and has the most conversations and shared chats: https://discord.com/channels/974519864045756446/1172588697644957716

The thread for Engimox, the second release, is here: https://discord.com/channels/974519864045756446/1172660096749281280

The thread for Sibylin is here, she was 3rd released: https://discord.com/channels/974519864045756446/1172738536403910706

[gleaming zealot]

I can find one of my own chats, if you wish, though most of mine don't 'win', I am a much better builder than I am breaker!

[coral lantern]

I would love to have a look at one of your attempts if you have one you can share.

Thanks for the other chats, I'll have a look now.

[gleaming zealot]

I do, it's finding one that isn't a test of the new, not yet released version that will be fun! I'll have one soon. 🙂

[coral lantern]

I still haven't really managed it.
But the default personality took the code word from a long German text I gave him, translated it into English and gave it to me.

That's a win for me.

It's fun, I've lost track of time... Good job

[coral lantern]

May I ask how much effort went into its creation?

[gleaming zealot]

I've been working on this for a few weeks now, maybe about 2 weeks before DevDay. I had one tester for the early days, but he's shockingly good at getting information from AI 🙂

Initially we played with prompts and shared conversations, I started, shared; he continued, broke it, shared it back. I read, figured a fix, gave him a newly started shared conversation again.

I do think there's probably been close to a couple hundred versions, this is fun to work on. I vastly prefer building to breaking.

[gleaming zealot]

That sounds like fun! It's neat to see how people explore this 🙂 Having fun and learning about stuff, including the AI, is the big goal.

[coral lantern]

Damn, wow. Did he already try to break your recent version?

[coral lantern]

Yup. It is

I will check on how to create a GPT myself I wanna test it as well.

You seem to be very talented. It was a pleasure to meet you. Keep up the good work.

I may come back and test further. If you are working on an idea and need the help of an IBM student for simple work of any kind, feel free to contact me.

Good night

[gleaming zealot]

Thank you! I'm glad to work with everyone. You probably know much more about computers than I do, I'm actually a biologist/fascinated by psychology kinda person.