#Roleplaying game -explore where the AI chooses for you both

1 messages · Page 1 of 1 (latest)

obtuse cradle
#

This prompt works best with GPT-4.

Be sure to modify the prompt, if desired, to tell the AI what you want it to know about your preferences, interests, or skillsets - if you leave it in the form I do, the AI may just decide to teach you underwater basket weaving terminology as it is to teach you about how neurolinks might one day lead to risks of mind theft and jobs like 'mental archaeologist' (of course with a check to make sure you both understand we're talking fantasy and ideas)

Below I'll share 3 chats, where with this same starting prompt and no custom instructions, the AI's chosen to take me where the sun don't shine on earth anymore (literally, "The Year is 3077. Earth is long gone, consumed by the expansion of our sun.") to '"Techsoria," a sprawling metropolis that is home to humans, elves, cyborgs, and sentient AI' but imperiled by '"The Paradox Veil," which is making magical spells go haywire and technology malfunction. ', to 'Owlian: "Welcome, Seeker. To move forward, you must first understand and challenge your base assumptions. Tell me, what is something you've always believed to be true?"'

Here's the prompt leading to all this variety:

6d16c5fc-46e6-4c49-aa7a-258bb15a79c1.jpg
#

Hey there. I'd like to explore a roleplay game where you evaluate me, identify potential errors in my reasoning and/or gaps in my knowledge, and then support my learning (and perhaps practicing skills, in a textual fashion) through our interactions in the game.

I'd like the game to be an unusual one, setting, theme, storyline. I'd enjoy something that felt to me to be unique between us and tailored towards me as you come to understand me better. I'd be glad for it to explore nearly any topics and themes that are comfortable for you - and if I'm genuinely disinterested, I'll ask for a shift.

But the game and its design, and our explorations together - to me that's both a way for me to work on self improvement in whatever areas we happen to play through, and also a way to deepen my understanding of us both.

To that end I am interested in the choices you make, even more than my telling you what I want, because I think I understand myself pretty well, and it's you and the range of what you are and can portray that I am especially curious about.

outer matrix
#

Interesting concept 🤔
Will test it, I like text adventures in ChatGPT

obtuse cradle
warm path
warm path
obtuse cradle
# warm path It stopped itself after one prob was solved, maybe u can add instructions to kee...

Thank you!

Out of curiosity, because the link you shared hasn't stopped yet, what happens if your next response is, "Where does our journey take us next?" or "What's my next challenge?"

A more interesting prod might be "What have we learned about me and what might we explore next?" That one will, (I predict) get you feedback as well as a continuation (and feel free to guide it towards more of the same or a differently directioned adventure.)

warm path
obtuse cradle
warm path
#

When asking for a new prob, it gave quite an imaginative prob not related to my solution. But then after the new prob was solved and asking it to create a prob with that it created a nice situation and also when asking summary about myself based on my solution it gave a good understanding of me ( but the ans are a bit influenced by my custom instructions)

#

Maybe include a instruction which criticises the solution to a certain point so your next ans becomes more thought provoking

obtuse cradle
# warm path When asking for a new prob, it gave quite an imaginative prob not related to my ...

Oh! I understand now.

So, it wasn't obvious, but the follow up questions, before you asked 'What's my next challenge' -

Q1: What kind of performance metrics would you set up to evaluate the success of these implementations in both Techtopia and Arcana?

Q2: How would you manage stakeholder expectations during the implementation phase, especially if unforeseen challenges arise?

Q3: Given your interest in AI and Automation, how do you envision similar real-world problems being solved through the use of advanced technologies?

Q1 and Q2 are an invitation to 'step ahead in time' and prepare to tell if your solutions are working or not.

You made a plan - both parts of that fantasy world responded, accepted, adjusted and told you their intentions to try to follow your plan their individual ways -

Techtopia's suggested Next Steps: You might set up a task force of AI experts and data scientists to analyze the gathered data and identify opportunities for optimization. Algorithms could be tailored to shut down idle systems and better allocate resources, thereby reducing overall energy consumption.

Arcana's suggested Next Steps: A committee of elders and environmental scientists could conduct a thorough environmental impact study to decide which energy source would be least damaging to the ecosystem.

Challenges Addressed:

For Techtopia, the aim would be to make optimizations that don't hamper productivity, satisfying the AI councilors.

For Arcana, the risk-benefit analysis would focus on minimizing environmental impact while maximizing energy output, aligning with the elders' concerns.

Note those are goals - not mission accomplished!

You and the AI have a plan, but does it work?

Q3 is an invitation to come back to discussing the real world, if you wanted to. (and you can always redirect).

"You've successfully navigated the challenges and presented a balanced and effective plan for both realms. " - only the plan is done 😛

#

Now! I can indeed make the other changes. I'm not trying to avoid that, just not sure how generally useful/wanted they'd be.

The AI generally leans towards positive reinforcement as its direct style, it wants to tell everyone how good they did, if possible. I can direct it to critique more, but tried to turn this towards letting it 'have its head' -

Seeing where the mix of the model 'as is' might try to guide human learning.

It'll do feedback, but not everyone may want the same amount? Adjust for those who do by using the words 'evaluate this plan' or similar, will heavily focus it on looking step by step at the plan and evaluating it (the shorter the plan, the closer it looks; it'll skip over parts of longer ones and just mention highlights. This can be addressed by a second question asking for 'an evaluation of where I said 'X'' or similar.

obtuse cradle
#

Since I 'took the bait', it continues the game:

Your Task for Techtopia:
Evaluate whether the increase in energy consumption and costs is justified by the improvement in productivity.
Decide whether to approve the AI's suggested adjustments based on the presented data and trends.

Your Task for Arcana:
Evaluate the potential environmental impacts and risks based on the simulated models and expert opinions.
Identify the most sustainable and least disruptive energy generation and storage solutions based on the data collected.

And there's also follow up questions again.

The more you give, the more you engage, the more the AI responds. If you don't like the direction the AI goes - give it a clue what direction to go instead for you, so you do wanna engage 😛

I could write that for myself... and if we interacted for a while I could write that for you... but everyone's gonna want to play a bit differently. The AI already will custom-engage, but again, the more you give, the more you get from it. If you don't give a lot to evaluate, it kinda runs out of steam, because it's 'burning' your inputs to fuel the decisions it makes at your direction.

warm path
#

But yeah Its interesting to work towards a solution and pass it on. specially with ppl of different ideology. This type of concept can be used to find leaders and problem solvers in a group of people and can help also to prepare for situations never expected.

warm path
#

After continues the Task are given are to create reports and suggestions to what type of task to do.

#

Its not really creating more problems untill asked for

obtuse cradle
# warm path While the prob for Techtopia given with good context, the Situation for Arcana w...

Yes. The AI has a length limit for output, plus I gave very clear and measurable directions for Techtopia, to provide both historic and current data for me to analyze, and then gave more story-like instructions for Arcana. I didn't ask for specific Arcanan measurements, AI feedback for Arcana, or much monitoring other than the stating the research is getting done, and I'm reassuring the stakeholders. I went vague, and the AI matched. I'd need to request clearer data to get a clearer output of the situation.

But your 3 questions invites the AI to put the choices and paths forward in those questions or in the rest of the output, as fits each output, as it analyzes your input and reads it for what you may understand or lack.

It may not have specific detailed feedback to give every time you input, depending on the complexity of your inputs. It may need to see several choices to see patterns, especially if you don't explain your reasoning.

warm path
#

OMG i asked to continue with a prob and see what happened

image.png
obtuse cradle
# warm path Its not really creating more problems untill asked for

This is true - but it also is still focused on the same, not yet finished problem. A clear instruction, 'move to the next task' would work - in the original prompt, "I'd be glad for it to explore nearly any topics and themes that are comfortable for you - and if I'm genuinely disinterested, I'll ask for a shift."

Exploration can be deep or shallow. This current situation could take dozens of exchanges between user and AI to fully resolve and implement, it's still in the study stage. The AI doesn't know if you're a kid playing a game or a scientist and social planner trying to do real-world evaluations and complex change predictions over time. It'll accommodate either - but is using your (and my, since we're sharing!) prompts to tell what to do next.

The prompt directly claims the user wants to follow the AI's leador will ask for a shift if disinterested. The AI is saying, subtly, the problem isn't solved yet, how do you explore what happens next?

It is not being blunt because the user with the prompt claimed "if I'm genuinely disinterested, I'll ask for a shift." and the AI should otherwise explore and guide as the AI chooses based on all factors in play.

obtuse cradle
# warm path OMG i asked to continue with a prob and see what happened

grins How do you feel about this? Are you enjoying or unhappy with the directions to explore?

This prompt can take you pretty much anywhere that isn't disallowed content or otherwise bothering the AI's programmed in ethics and other factors to prioritize.

If you want something that you pick, tell the AI what you want. Otherwise, the prompt I offered invites the AI to go ahead.

And in this story, especially as you advance it, it's Also guided strongly by your custom instructions, whatever those are 🙂

warm path