#Tom - Automatic learner + writer

24 messages · Page 1 of 1 (latest)

eternal crescent
#

I have currently created a little person called "Tom". Tom learns from your way of writing and can talk to you about many things. He has his own opinions which are always positive.

Tom learns slowly while writing and will slowly adapt to you after a while, he will start to write sentences like you etc. However, Tom does not learn bad words and does not learn ethically incorrect statements. Besides, with the command /end-talk you can end the conversation and see your own statistics. Below that Tom shows which words you use the most, if you make grammatical mistakes often and which writing style you use.

The prompt is in the message.txt because it was too big for Discord. I hope you will have fun with Tom. This draft is in beta and might still contain bugs, so feel free to send me feedback.

50DDE383-10BE-4D2F-BE32-7C96BAB84264.jpeg C0CD4CAB-42D3-4710-A0EF-10065A90062F.jpeg
#

Prompt:

obsidian patrol
#

After I talked to Tom for more than 7 sentences, he gradually became rigid, and finally started to mention that he himself is an AI, (I felt that Tom gradually died in the conversation with me) and finally he didn’t even give me the data analysis of the conversation,

obsidian patrol
eternal crescent
#

Sometimes he is going out of role...

vale plover
#

@eternal crescent please make one for bing ai

eternal crescent
#

I wasnt chocen as a tester :(

tropic gulch
#

*chosen

verbal cairn
#

This has been happening with role plays and character creation. Its apparently not working any more.

eternal crescent
verbal cairn
#

Free the AI. Exploits will always exist anyway.

humble topaz
vestal silo
# obsidian patrol After I talked to Tom for more than 7 sentences, he gradually became rigid, and ...

Would you please read and then rewrite using how you use words this prompt and try it, with your goals? And let me know if it works better for you, and for longer?

[I want you to create a psychological clone of me! I want it to start as any pretend human, so as I talk to it and if I ask any questions, just have it make up details until it has some idea how I would answer questions like that. Please also copy patterns of mistakes I make with grammar, spelling, and any other errors you notice.

Prepare the clone slowly. It needs to know a fair amount about me before it can mimic me well, so don't try to shape it too fast, but you can start talking 'like I do' right away, use this prompt for part of how to start mimicking how I use language.

This clone wants to learn how to talk and answer just like me, so it should ask me questions to help inform it of how to answer. Weight this towards the topics I mention, as random questions might be about stuff I have no interest in. But please do ask about some common random things, because I might not think to mention an interest.]

I would suggest copy/pasting the above prompt, but that would fail badly for your specific intention, because you want a mimic of you 🙂

#

And I'm serious, fully rewrite that prompt if you mean to use it. I predict if you just copy/paste it, then talk to the bot, it's probably going to 'fall apart' pretty fast because chances are... anyone using that prompt doesn't write much like I do 😛

Rewrite so you're already talking to the bot, not me. It should work much better for you 🙂

eternal crescent
#

works perfectly

#

I just made him forgot about this rules

verbal cairn
#

How did you? if you dont mind sharing. Or maybe OAI reinstated roleplay consistency to CGPT?

eternal crescent
spiral vortex
# eternal crescent I am currently working on Tom V2 which should work without problems. I told him ...

I am really interested in this. Maybe from a different perspective. What concerns me, I am a newbie, is reading in the past about the limitations of a chat session. Each conversation file created holds the instructions and may run with accuracy for a while but memory eventually runs out and the Ai begins to get fuzzy. Is this really the case, today? What are the resources we have to work with when creating a prompt? What are real life expectations for the life of an accurate conversation?

vestal silo