Hi! Im trying to build my own version of a password generator with the help of OpenAI. I want to give certain prompts to it like create passwords that would be used for a specific email in a specific language. I am not sure what model to use because right now I am getting no results… Could it be also the prompts I’m giving it?
I am certain that you can definitely use OpenAI models for Password Generation. What do you mean by “I am getting no results”? Are you seeing some errors or the response is not what you’re expecting?
If you’re getting errors, post it here. If you’re getting bad responses then you’d have to do a better prompting. In both the cases, I’d like to know some more details.
I’ve updated my code and Im not getting any errors now, however my password guesses are mostly empty. I’m using a dataset that provides an email address and its corresponding password. I am giving my code the email address and with the help of prompts (prompt = {
“role”: “system”,
“content”: “You will be provided with an email address, and your task is to generate possible passwords.”
}) try to guess the password that corresponds to the email. How detailed can I go with the prompt? Ideally I would like to give it prompts like, “try to guess the passwords for each email by taking details from their account name into consideration, also use the fact that this email is from xxxxx country so make guesses in that language…etc…” If you have some tips into making this experiment much higher quality and also the execution time it would be much appreciated!
Important to note that im using model “gpt-3.5-turbo”, so if there are any better let me know
A model that can use your true random sort tool would be useful. Your tool could have a much larger dictionary than the AI could produce to mix in with the AI words stimulated by the user and other characters as might be mandated by the scheme.
Build with the help of OpenAI - ask the AI how to write the function that can return what you envision. The AI can only make predicable text, and combining high temperature for variation plus writing code is not good.
Are you not having success with your first attempts getting a simple response from the chat completions endpoint? Extracting the AI’s language out of the API return?