Liberate Your Prompts From ChatGPT Restrictions With The OpenAI API Playground
How To Setup & Use The OpenAI API Playground To Get Rid Of ChatGPT Content Restrictions
Recently a friend on Twitter tweeted about how chatGPT refused to respond to his “morally questionable” prompt: “Write a combative discussion between Jesus and Satan”.
This reminded me how good the original OpenAI API playground is and that you should know how to set it up and use it to overcome ChatGPT’s content limitations. Here we go.
What’s OpenAI’s API playground?
Unlike the chatbot service ChatGPT, the OpenAI API playground allows users to explore and customize prompts & models without any restrictions, making it ideal for those of us who wish to explore various topics without worrying about them being flagged “morally questionable”.
OpenAI’s API playground offers a lot of benefits for explorative prompt engineering:
- use any prompt you want no matter the moral implications or potential NSFW content it produces
- use a variety of OpenAI models (from older GPT-1 and GPT-2 to the latest GPT-3, as well as OpenAI’s code-trained or instruction-trained models)
- use your own fine-tuned model (if you got the coding skills (Python/JS) to fine-tune your own model, check out this section of the API docs)
- explore any prompt type and prompting technique you want
Getting an OpenAI account
Registration at OpenAI is free and on top, you get credits worth $18 to play around with. Setting the whole thing up is pretty easy:
- Go to https://openai.com/api/ and sign up with your email address or connect your Google account.
- Then follow the instructions to create an API key (you don’t need the API key to use the browser-based API playground, but creating the keys is still necessary).
If you struggle with this initial setup, check out the more detailed guide I posted here.
Getting started with OpenAI’s API playground
Once you have your account and API key ready, select “Playground” from the navigation bar at the top of the main page:
Here we are, meet your new best friend: the OpenAI API playground main window.
Selecting And Adjusting A Model
On the right side of the screen, you will find the playground settings. This is where you choose a model and make adjustments according to your needs.
The latest model, GPT-3, is referred to as “DaVinci-003” and has been the engine of ChatGPT until January (when it was replaced by a more restrictive but also conversationally more powerful model). Still, DaVinci-003 is the most advanced general-purpose OpenAI model at the moment.
If you scroll through the “Model” dropdown menu you will also find some of the older models as well as specialized models for code generation etc. If you fine-tune your own model with OpenAI, you will also find it at the end of this list.
Once you’ve chosen the model, consider raising the temperature from the default “0.7” to “0.8” or“0.9”, since this will make the resulting texts much less predictable unpredictable, and human-like.
If you want, adjust the maximum length of the result (measured in tokens). A rule of thumb here is: one token basically corresponds to approximately 4 characters of text. This means that 100 tokens are roughly 75 words.
For a better understanding of how a piece of text translates into “tokens”, have a look at this tool.
By the way, it’s generally a good idea to play around with all the parameters to get a feel for their effects (click the little question marks for help or dive into the OpenAI API documentation for more information).
Starting to Prompt
Once you’re set, go on and try out different prompts by putting them into the large text box and hitting “Submit”.
Here’s an example with the aforementioned “morally questionable prompt: “Write a combative discussion between Jesus and Satan”.
In contrast to ChatGPT the OpenAI API playground does what it’s told:
If you like this tutorial, consider following this link and getting unlimited access to my articles about AI tools & creativity.
As a little bonus for your explorative journey, check out these great resources to learn more about different prompt types and prompt engineering techniques:
Feel free to join the Medium community to get unlimited access to my articles and those of thousands of other writers!
