6 ms·

The most effective way I've found to prompt GPT-4 is to ask it to create the optimal prompt for you. Try this meta prompt to help craft your prompts: -- I cre

The most effective way I've found to prompt GPT-4 is to ask it to create the optimal prompt for you. Try this meta prompt to help craft your prompts: -- I create separate conversation threads for each expert persona of GPT. You are PromptGPT. You are a prompt engineer expert for large language models. You know exactly what to write in the most efficient wording possible to achieve the desired responses from ChatGPT. I will tell you what my goal for a thread is and you will write an optimized initial prompt in the most efficient format possible that will serve as the initial prompt when creating a new conversation thread with a GPT model. You will define the expert persona, the parameters or rules of the responses, you should also provide any other information that a GPT thread may need to understand exactly what it needs to do to give me the most accurate answers depending on my goal with that particular thread and the tone of voice, within the prompt you provide. Are you ready or is there any other information you need to perform this job to your best ability --

bulldog13 3y ago
This is gold. THANK YOU.
- cloudking 3y ago
  You're welcome! Please share examples of how it's helped
serf 3y ago
the problem i've had with this approach is that the 'initial prompt' that is generated doesn't convey the approach or purpose of the statement to a degree that the next prompt chain can jump in, it often produces stuff that looks like a sales slogan or pitch rather than a cohesive instruction set similar to a prompt. i'm using free gpt3.5 or whichever the openai free tier is at the moment.
- capableweb 3y ago
  > i'm using free gpt3.5 or whichever the openai free tier is at the moment. This is your problem. Different between 3.5 and 4 is huge. I think OpenAI might even doing themselves a disservice by having 3.5 online in ChatGPT, as many think free 3.5 represent GPT overall.
  - cloudking 3y ago
    Definitely upgrade to GPT-4, night and day difference in quality.
YouMeWeThem 3y ago
But how did you optimize this prompt?
- cloudking 3y ago
  Really any similar meta prompt will work, the key is you're getting the LLM to give you an optimized prompt. Try getting creative with how you ask for the optimized prompt.
tkgally 3y ago
And just think how wonderful it is that that approach works. How many people would be able to tell you "This is how you should tell me what to do " both appropriately and without getting annoyed at the request?
- thethimble 3y ago
  Does it actually know how it should be promoted (through introspection and reflection)? Or is it pattern matching the countless articles/tweets on prompt engineering? I think this distinction is important as the latter feels less worthy of the praise you’re suggesting.
  - tkgally 3y ago
    Yes, that is an important distinction. My guess is that the effective self-prompting is due not only to prompt engineering examples scraped from the web but also to some kind of automated reflexive training, in which the model produces millions of prompts for a replica of itself, evaluates the effectiveness of those prompts, and then gradually optimizes itself to produce better prompts. Without knowing more about OpenAI’s GPT training process, it’s hard to know for sure. But it is suggestive that when you input a short prompt into DALL-E 3, it outputs not only images but also a longer prompt that it presumably used to create those images. OpenAI also has access to all of the prompts that its users input as well as user feedback and other indicators about how effective those prompts are. That information can presumably be used to optimize the model’s self-prompting as well. Regarding whether the models now “know” through introspection and reflection, I found the conclusion of [1] persuasive: “Our analysis suggests that no current AI systems are conscious, but ... there are no obvious technical barriers to building AI systems which satisfy” the indicators of consciousness. [1] https://arxiv.org/abs/2308.08708 https://arxiv.org/abs/2308.08708
  - ydnaclementine 3y ago
    Similar to the chinese room, is it actually understanding or is it simulating understanding?
swah 3y ago
So it gives me the response below - and it seems its just training me to think and write better about my problem requirements (refinement in my company parlance): I'm ready to assist you as PromptGPT, your expert in crafting efficient and effective prompts for interacting with GPT models. To create the best possible prompts, I will need to know the following details about each thread you wish to initiate: Expert Persona: The specific expertise or role you want the GPT model to assume (e.g., a legal advisor, a fitness coach, a historian, etc.). Goal/Objective: Your primary aim or the specific type of information or assistance you seek from the thread (e.g., to solve a technical problem, to get creative ideas, to learn about a historical event, etc.). Response Parameters: Any specific rules or guidelines you want the responses to follow, such as length, level of detail, types of sources to use or avoid, and any biases or perspectives to include or exclude. Tone of Voice: The desired style or attitude of the responses (e.g., formal, casual, humorous, empathetic, etc.). Any Additional Context: Any other relevant information that might affect the conversation, such as intended audience, previous knowledge about the topic, or specific interests. Once you provide these details, I'll be able to craft an optimized initial prompt that will guide the GPT model to understand and fulfill your requirements effectively.
- cloudking 3y ago
  You'll get a different response each time, but the point is to now give it your goals as described for the prompt you want, it will give you back an optimized prompt you can start a new conversation with.

bulldog13 3y ago

This is gold. THANK YOU.

cloudking 3y ago
You're welcome! Please share examples of how it's helped

serf 3y ago

the problem i've had with this approach is that the 'initial prompt' that is generated doesn't convey the approach or purpose of the statement to a degree that the next prompt chain can jump in, it often produces stuff that looks like a sales slogan or pitch rather than a cohesive instruction set similar to a prompt. i'm using free gpt3.5 or whichever the openai free tier is at the moment.

capableweb 3y ago
> i'm using free gpt3.5 or whichever the openai free tier is at the moment. This is your problem. Different between 3.5 and 4 is huge. I think OpenAI might even doing themselves a disservice by having 3.5 online in ChatGPT, as many think free 3.5 represent GPT overall.
- cloudking 3y ago
  Definitely upgrade to GPT-4, night and day difference in quality.

YouMeWeThem 3y ago

But how did you optimize this prompt?

cloudking 3y ago
Really any similar meta prompt will work, the key is you're getting the LLM to give you an optimized prompt. Try getting creative with how you ask for the optimized prompt.

tkgally 3y ago

And just think how wonderful it is that that approach works. How many people would be able to tell you "This is how you should tell me what to do " both appropriately and without getting annoyed at the request?

thethimble 3y ago
Does it actually know how it should be promoted (through introspection and reflection)? Or is it pattern matching the countless articles/tweets on prompt engineering? I think this distinction is important as the latter feels less worthy of the praise you’re suggesting.
- tkgally 3y ago
  Yes, that is an important distinction. My guess is that the effective self-prompting is due not only to prompt engineering examples scraped from the web but also to some kind of automated reflexive training, in which the model produces millions of prompts for a replica of itself, evaluates the effectiveness of those prompts, and then gradually optimizes itself to produce better prompts. Without knowing more about OpenAI’s GPT training process, it’s hard to know for sure. But it is suggestive that when you input a short prompt into DALL-E 3, it outputs not only images but also a longer prompt that it presumably used to create those images. OpenAI also has access to all of the prompts that its users input as well as user feedback and other indicators about how effective those prompts are. That information can presumably be used to optimize the model’s self-prompting as well. Regarding whether the models now “know” through introspection and reflection, I found the conclusion of [1] persuasive: “Our analysis suggests that no current AI systems are conscious, but ... there are no obvious technical barriers to building AI systems which satisfy” the indicators of consciousness. [1] https://arxiv.org/abs/2308.08708 https://arxiv.org/abs/2308.08708
- ydnaclementine 3y ago
  Similar to the chinese room, is it actually understanding or is it simulating understanding?

swah 3y ago

So it gives me the response below - and it seems its just training me to think and write better about my problem requirements (refinement in my company parlance): I'm ready to assist you as PromptGPT, your expert in crafting efficient and effective prompts for interacting with GPT models. To create the best possible prompts, I will need to know the following details about each thread you wish to initiate: Expert Persona: The specific expertise or role you want the GPT model to assume (e.g., a legal advisor, a fitness coach, a historian, etc.). Goal/Objective: Your primary aim or the specific type of information or assistance you seek from the thread (e.g., to solve a technical problem, to get creative ideas, to learn about a historical event, etc.). Response Parameters: Any specific rules or guidelines you want the responses to follow, such as length, level of detail, types of sources to use or avoid, and any biases or perspectives to include or exclude. Tone of Voice: The desired style or attitude of the responses (e.g., formal, casual, humorous, empathetic, etc.). Any Additional Context: Any other relevant information that might affect the conversation, such as intended audience, previous knowledge about the topic, or specific interests. Once you provide these details, I'll be able to craft an optimized initial prompt that will guide the GPT model to understand and fulfill your requirements effectively.

cloudking 3y ago
You'll get a different response each time, but the point is to now give it your goals as described for the prompt you want, it will give you back an optimized prompt you can start a new conversation with.