Summary

OpenAI Playground is a tool which provides an environment for people to experiment with reinforcement learning (RL) algorithms. With the OpenAI Playground, users can design an AI agent that learns to effectively play classic Atari titles, as well as custom game environments, create a chatbot and other things. This interface offers a not so streamlined, but easy and way faster possibility to check out things than the ChatGPT site.

The package documentation can be found here: https://irudnyts.github.io/openai/index.html.

Setting up things:

In order to get access to the OpenAI API, you need to optain an API key first. Therefore, please create a new account here, or if applicable log in with your already existing one/and alternative provider (e.g. google).
Once you signed up and are logged in, please open this page. Which is the place for your API keys. If you have none yet, create a new one, via the drop down arrow make the whole key visible and copy it into your clipboard.

Next you need to decide on which level you want to store this API key:

Table listing recomended places to store the OPENAI_API_KEY variable.
Environment level	usethis call	description
project	`usethis::edit_r_environ(scope = "project")`	The specific Rstudio project needs to opened to query, as the environment variable is stored in a project related manner.
global	`usethis::edit_r_environ()`	With this setting this environment variable is available in all your Rstudio Sessions. (with the respective user).

Add the OPENAI_API_KEY to your environment.

if (!require(usethis))
    install.packages("usethis")

usethis::edit_r_environ() # global
# OR # 
usethis::edit_r_environ(scope = "project") # project specific

In the file that opens now add your token with the following syntax, replacing the xx-es by your personal token, which you can OPENAI_API_KEY = 'xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx'

Application

Available models

First we load the library, and then call for all available models, due to easier visibility only parts of the returned data are shown here.

library(openai)

list_models()

id	object	created	owned_by
babbage	model	1649358449	openai
davinci	model	1649359874	openai
gpt-3.5-turbo-0301	model	1677649963	openai
text-davinci-003	model	1669599635	openai-internal
babbage-code-search-code	model	1651172509	openai-dev
text-similarity-babbage-001	model	1651172505	openai-dev
text-davinci-001	model	1649364042	openai
ada	model	1649357491	openai
curie-instruct-beta	model	1649364042	openai
babbage-code-search-text	model	1651172509	openai-dev
babbage-similarity	model	1651172505	openai-dev
gpt-3.5-turbo	model	1677610602	openai
code-davinci-002	model	1649880485	openai
code-search-babbage-text-001	model	1651172507	openai-dev
text-embedding-ada-002	model	1671217299	openai-internal
code-cushman-001	model	1656081837	openai
whisper-1	model	1677532384	openai-internal
code-search-babbage-code-001	model	1651172507	openai-dev
audio-transcribe-deprecated	model	1674776185	openai-internal
text-ada-001	model	1649364042	openai

ask a respective model for code completion

This can be easily fine tuned before at the openAI website.

text <- openai::create_completion(
    model = "text-davinci-003",
    prompt = "Write a little story about a luck little programmer which was at the right time at the right spot to discover something novel",max_tokens = 4000
)

knitr::raw_html(text$choices$text)

Once upon a time there was a young programmer, who was trying to make a breakthrough in his career by working towards creating something novel. Each day he worked long and hard, and eventually decided to take a break from his coding and get out into the world. So he took a walk in the park and enjoyed the sound of birds chirping around him and the cool breeze. As he went further down the path, he stumbled across an old laptop with an open source code on the screen. Being very curious, he decided to take a look and much to his surprise, he noticed that it was a code of a new game. He didn't know it then but it would eventually become a huge success! The young programmer was overjoyed as he quickly got to work to implement the game. His task became easier as he increased the speed of the game's functionality and added more levels and features. Eventually, after several months of testing and tweaking, he was able to officially launch the game and it was a huge hit! With the newfound success, the luck little programmer became an overnight celebrity. People all over the world were talking about his success and appreciating the hard work that he put in to create the novel game. From that point on, the programmer was very successful in his career path and became an inspiration to many young programmers. His story proves that if you work hard and stay lucky, anything is possible.

Such a nice hard working bioinformatician ;)

Thanks for reading until here!

Regarding promt design I can recomend only to work through the following link yourself, as I haven’t gotten any further yet openAI Guide on text completion.

For testing how authentic your text is, you could try out the openAI text classifier.

And if you by chance know a way to get the returned html text nicely formated acording to the breaks, please feel free to leave me a comment.