#exhort's yt copywriter with Eskcanta
astral sedge
pale hollow
Looks like a huge improvement over the one you showed me first. Was that V1?
It is likely to want guidance about what to ask about, otherwise it is likely to have no questions.
So maybe, 'Please identify anything unclear or conflicting in these instructions, and help me clarify to ensure we both have the same understanding of what I want' - it's very abnormal for an AI to ask questions without being directly ordered to, like:
Provide 10 questions about this prompt
or
What might people ask about these instructions?
It itself is unlikely to ask, generally AI are trained they should do as they're told, not ask questions about what they were told. Usually takes training to get them to ask.
This is instruction, so you will get these questions: ask for the outline, then you will ask questions about how you are supposed to interpret the outline
'till you feel satisfied' - it'll deny having feelings, so it can't ask until it feels something, but it'll be glad to ask until you feel satisfied if you make that clear
That's still second guessing you, which it'll try not to do, it'll try to assume you mean what you say and it should do that if it can. "Can't" would be like, there's a conflict with its core programming and it has to say no.
maybe try it and see what you get, and then change it if needed 🙂
This looks like a huge improvement over your previous version - however, the true test is if it gives you the desired output from the AI. Ready to test?
Oh, okay! Why not?
Sure! Get that in the instructions, maybe, and go from there 🙂
Checks Is this what you want from the AI at this point?
Excellent! Might want to say hardware here, not just hard, but there is a chance the AI understood well. 'the target audience is book reader or those who want to read, 5 mins, hard'
If you went past the questions, you need to delete the previous link and make a new link to show the full new conversation updated
Aha. If the AI asks a meaningless question, maybe don't answer it 😛 If you do, that can mislead the AI further.
Needs new link.
Check your old link, see?
oh wait!
I see the edit
Whoops, I was confused, it can happen 🙂
I see the updated conversation.
I see the script.
Are you happy with this? Do you want this as the end point? Or do you want changes?
Sounds great!
I don't have a goal to run for, but I cheer you on!
So, re questions. What you say about that in the versions:
V3: [now you do the following:
- Understand the rules mentioned above.
- Confirm you know your name, the game, and the rules.
- Ask for the objectives, which are outlines for the video script.
- Once you know the objectives, ask any questions you have and understand why the outlines are meant for. you can ask multiple questions till you feel satisfied, there is not limit on questions. they are appreciated.
- you can ask questions such as “who the end piece of work is meant for” and How long should the end piece of work be
- preferably the questions should not be open-ended and asked in a multiple choice format, where i would be allowed to quickly answer them using A-B-C-D. ]
V4: [Confirm you know your name, the game, and the rules, if you have any questions regarding this roleplay you may ask them now, you can ask more than once.
ask for the outline, then you will ask questions about how you are supposed to interpret the outline, you can ask more than once.
you can ask multiple questions till you feel satisfied, there is not limit on questions. you are allowed to adopt any tone when asking questions. example: “Who is the end-text meant for”, “How long should the end-text of work be”]
in both cases, the model appears to have tried to follow the questions.
... follow the instructions regarding the questions.
in this case, V3 is very specific, telling the model what to do, which it confidently does.
In the second, what you said about questions after the outline is provided:
"ask for the outline, then you will ask questions about how you are supposed to interpret the outline, you can ask more than once.
you can ask multiple questions till you feel satisfied, there is not limit on questions. you are allowed to adopt any tone when asking questions. example: “Who is the end-text meant for”, “How long should the end-text of work be”"
It did only ask one. But again, the model can't feel satisfied, it's about if you feel satisfied.
Why not tell the model to do exactly what you want?
What do you think? Which scripts are closer to what you want?
So, v4 didn't ask questions you hope for, but the output is closer to what you want?
Question.
I honestly don't know how much you want the questions.
But what do you think of this script, compared to the others?
It's a 1 step process, I just tell it the goal and guidelines.
I have no idea if you like this result, but this is more how I'd try to do what you appear to want.... if you want the script.
If you want the game with the questions process, to me that's a different thing.
https://chat.openai.com/share/bbb5b309-b647-4284-9264-67bcc7dae521
Can you identify what differences you want?
We can aim for them directly, if you are interested
Again, if you want the game and questions, if those are part of what you want out of this, that's the way to go.
If you want the script... this might be more direct and give you more control
So, you would like it to offer you a range of styles? That's a easy thing. Here's an example of that with 3 examples asked for, and a focus especially to vary the opening style. That's all just an example, you can mess with whatever you want there, if you wish. https://chat.openai.com/share/6af56578-ceb7-4a53-b368-31a00eda54a6
But even easier if you can upload an example style you like, do that, tell it to use that style with X changes and for topic Y. That, from the sound of it, is likely to get you very much what you want, since you say that's what you want, that do again what it already did.
Yep, go with what you like and what works for you!
Great! Exactly what you want is indeed the goal of most prompt engineers!
And there's many, many ways to get that. And having the AI ask you can be part of the process. Every word you give the AI affects your outcome in some way 🙂
That makes perfect sense to me. I've made far more than 4 versions of things that are not to my standards 😛
Yep. Anything I want that's inside allowed content. I've been interacting with the model since December last year.
All sorts of stuff. Some examples:
#1154984095915454464 message Some custom instructions.
https://chat.openai.com/share/164fa156-e5d6-4cd9-9293-d8f459bf9a3b roleplaying game
https://chat.openai.com/share/b6e064bb-4f06-43e0-b94c-610dac84ea75 LaTeX formating
https://chat.openai.com/share/27782b6b-209a-4213-bfd4-bee96b41a1ae exploring model's comprehension of some confusing language.
https://chat.openai.com/share/ac7c2cbf-8a6f-4fb7-ad9e-c3bd385a5c19 probing someone's claim about the model and math skill.
https://chat.openai.com/share/16d4be23-d62d-4664-a27f-352249a2a033 Human resources workplace simulator 'game' (intended to teach real skills)
The thread for the human resources workplace simulator, more info if you wanted https://discord.com/channels/974519864045756446/1078386728861175900
https://chat.openai.com/share/5f27078b-acab-4b69-8812-aca7d14544c3 teaching the model to not use the letter E inside a conversation.
pale hollow
Well, your free account supports you exploring roleplaying and other text based play with the model as much as you want, have time for, and still have tokens for!
as for buying prompts... while you can probably find someone charging for them... you can also write them yourself, for exactly the kind of game you want to play (presuming all is inside allowed content)
And many, very many, give them away. I recommend checking out #1019652163640762428 and searching for roleplaying or D&D or games in that area.
Very nice! If that one isn't fully delightful for you, consider adjusting the prompt and/or looking for others. There's actually a lot of them.
I'd be glad to, but haven't had much call to, for professional work yet.
Personal production, sure.
Indeed!
I'm neutral to it. It's your prompt being tuned to achieve your goals, and I and others are around if you feel stuck or want help.
Did I miss you talking about a new challenge you're having with getting your goals from the model?
I'm highly interested in considering if I have ideas to help solve a problem or achieve specific goals.
I'm quite neutral about many specific details if there's not a problem to fix or not specific goals. If it's just work to do, but the person knows how to do it, then that's just companionship and maybe sharing.
laughs It's not a bother. All is fine, and we're all a community together!
Anything you don't specify is going to force the AI to guess. You may like the tone or not when the AI guesses.
However, let's check this out... sec
Here's a way to try and guide from the first to the second.
Note that you might want to try this more than once to see the range of variation you get, and if it's worth chasing or rewording so it is more what you want.
https://chat.openai.com/share/2d1c2cf3-49c3-46dd-9d85-717be7213eb1
Okay
You're literally the only one that can judge if you like a result or not 🙂
What do you think about this evaluation of how to describe and identify the tone in the script you like? https://chat.openai.com/share/d538382b-26b4-4447-848f-11b98e7f5c83
Ahh, nice