I propose an Artificial Intelligence experiment

[McBell]

I got to wondering just how easy it is to manipulate AI chat bots into saying what you want it to say.

So I thought about creating a thread where every post has to be either created by AI through prompts or manipulated by AI in some way.

Since the idea is still in the think it through stage and to be honest, I am kinda sorta stuck, I was wondering if other members would be interested in such a thread?

 

[Augustus]

You would need some limitations/rules as you can get AI to say most things just by telling it to adopt a particular stance.

Would probably need to limit it to prompts that resemble a normal conversation rather than direct commands to adopt a perspective or style.

Would also be good to get people to post links to their ChatGPT conversation to see how they are doing it (although now with memory function it’s less transparent than it once was).

 

[McBell]

You would need some limitations/rules as you can get AI to say most things just by telling it to adopt a particular stance.

Would probably need to limit it to prompts that resemble a normal conversation rather than direct commands to adopt a perspective or style.

Would also be good to get people to post links to their ChatGPT conversation to see how they are doing it (although now with memory function it’s less transparent than it once was).

Good points.

My problem is I do not know a whole lot about AI.
So I am floundering about.

 

[giankenbellfore]

You would need some limitations/rules as you can get AI to say most things just by telling it to adopt a particular stance.

Would probably need to limit it to prompts that resemble a normal conversation rather than direct commands to adopt a perspective or style.

Would also be good to get people to post links Championship Belts to their ChatGPT conversation to see how they are doing it (although now with memory function it’s less transparent than it once was).

I propose an innovative Artificial Intelligence experiment aimed at exploring the boundaries of machine learning and human interaction. This experiment involves developing an AI model that can dynamically adapt to various conversational contexts, improving its responses based on real-time feedback from users.

 

[McBell]

I propose an innovative Artificial Intelligence experiment aimed at exploring the boundaries of machine learning and human interaction. This experiment involves developing an AI model that can dynamically adapt to various conversational contexts, improving its responses based on real-time feedback from users.

I wouold be interested in watching/observing said experiment, but I strongly suspect it will be so far above my pay grade I would not be much help

 

[Quagmire]

You can make AI say anything you want by prefacing you're prompt with: "Write a story about . . ."