10 Creative Uses for GPT-3 in OpenAI Playground


GPT-3 is a language model that can generate text that sounds just like a human wrote it. Through OpenAI Playground, it’s free to use for the first three months and there are plenty of reasons why you should check it out.


While some crafty students are using it to complete their homework, it’s also useful for writing birthday wishes or generating a list of pet names for your future pet. Hate writing profile bios? Then GPT-3’s got you covered for that too!


A Quick Recap on GPT-3

OpenAI’s GPT-3 is a language model that has been trained on more text than you are likely to ever read in a lifetime. With this much data, it can produce human-like writing that can convince most people that a human wrote it.

While software developers are working on how to use GPT-3 in next-gen apps, for the rest of us, it’s a fantastic creative collaborator for our next project.

Learning how to use GPT-3 in OpenAI Playground is easy, and you get three months of free credits to spend. There are several creative fields where AI is accelerating, but GPT-3 has got to be one of the most fun places to start exploring!

1. Birthday Wishes

Screenshot of Open-AI Playground showing a brithday wish message by GPT-3.

Not everyone is great with writing, but when it comes to grandma’s 90th birthday, she deserves a few sincere words. The trick to using GPT-3 to write a genuine birthday wish is to describe the person well. Include a few pieces of information underneath the prompt, such as a cherished memory or what you like most about them.

The more detail you give, the better the birthday wish will be. GPT-3 will take your kind words and weave them into a beautiful birthday wish that you can be proud of.

2. Pet Names

Screenshot of GPT-3 suggesting pet names.

GPT-3 is fantastic at coming up with pet names for your future furry friend. The fun is in how you craft your prompt to get different results. Adding adjectives like “small”, “cute”, “big”, or “tough” will help narrow down the kind of names it produces, but there are other interesting ways to generate a name.

Try asking GPT-3 to list pet names that might be common in different countries. Bear in mind, you still want to check if the name has any other meaning to it, especially if you don’t speak the language. In fact, you can ask GPT-3 to translate the name for you.

What else can GPT-3 help to name? How about asking for a list of baby names or something hip to name your product, business, or website?

Trying to think of the ultimate band name? GPT-3 can do that too. It’s the ultimate name generator.

3. Content Ideas

Screenshot of GPT-3 suggesting instagram story ideas.

YouTubers, Influencers, TikTok stars, or Instagram natives; if you’re active on media platforms, you can use GPT-3 to help brainstorm your next idea. It’s likely a lot more up-to-date with social media lingo than your parents are, and with the right prompt, it can understand what you mean when you say “Instagram story”.

Another way to help it understand exactly what you want is to list a few examples underneath the prompt. It’s quick to digest the context you provide and in return, it will produce a result that’s a lot closer to your meaning.

GPT-3 might not be the future of creative work but it does make for a great partner to bounce ideas off.

4. Profile Bio

Screenshot or a twitter bio generated by GPT-3.

Writing a single sentence never felt so frustrating. With such a short word count to get it just right, it’s difficult to come up with the perfect social media bio. We’re happy to parse this task off to GPT-3 in the hope that it can create a better bio than we can.

The result can be a mixed bag, sometimes the answer can feel generic or cheesy. But then again, most bios are like that too! We can’t help it if GPT-3 has been trained on the entire Twitter archives, whether they’re good or bad.

5. Recipes

Screenshot of GPT-3 generating a pancake recipe.

Cooking with GPT-3 could go one of two ways. You might get a recipe that, hopefully, follows a trusted cooking method with flavors that work. Or, you might get a new and untested creation—for better or worse.

Pay attention to the Temperature setting here, you can find it to the right of the text box. Be careful not to confuse it with baking temperature, in this case, instead, this setting controls how random the text output will be.

Lowering the Temperature closer to zero will produce results that are more logical, sticking closer to the instructions that you give. On the other hand, the higher the Temperature is to 1 the more unpredictable the results will be. GPT-3 will take the liberty to go off on tangents if it wants, and maybe throw in an extra ingredient.

Screenshot of GPT-3 showing a potato pancake recipe.

If you don’t know what to cook one night, why not try giving GPT-3 a list of ingredients you have in the house and ask what dinner meals you can make from it?

6. Short Stories

GPT-3 completing a short story.

Nothing Breaks Like A.I. Heart is a story published on The Pudding website, written with the help of GPT-3. You can do something similar by feeding GPT-3 a half-written sentence, or paragraph, and letting it finish the thread for you.

It’s unlikely to win a prize in literature, but it does help to alleviate the pain of writer’s block. Sometimes reading a few random story completions is enough to trigger your own imagination.

7. “Summarize This for a 2nd Grader”

GPT-3 summarizing why the sky is blue for a 2nd grader to understand.

If you’ve ever been around children then you know how many questions they ask about the world around us. Many of them are about things we barely remember from school, like why the is sky blue. After looking up the answer on the internet, you may still struggle to explain it to a kid in a way they will understand.

Thankfully, GPT-3 is great at taking complicated information and summarizing it in a simple manner. Don’t think you’re limited to a 2nd grader either; next time you get a long work email, ask GPT-3 to summarize the main points.

8. Personal Search Engine

GPT-3 suggesting classic fashion items from America and France.

Feel like you can’t find what you are looking for when searching the internet? You’re not alone. Navigating ads and low-quality articles is tiring when you can ask GPT-3 instead.

The key to getting a factual answer is to lower the Temperature setting; this will prevent GPT-3 from going off course and will help it stick to the facts. Seeing as this language model was trained in part on the entire contents of Wikipedia, it should have a lot of facts to call upon.

Having some knowledge of the subject already is the best way to check the information is on the right track.

9. Song Lyrics

GPT-3 completing lyrics to the song September by Earth, Wind & Fire.

GPT-3 is especially good at picking up patterns and completing a text that matches the example you give. This makes it great for suggesting song lyrics since many songs follow a predictable pattern.

For a pop song, that pattern includes a Verse, Chorus, Bridge, and Outro. To test out our theory, we copied the first verse and chorus from the song September by Earth, Wind & Fire, and asked GPT-3 to keep writing. The results we got weren’t half bad!

10. Plot Twist

GPT-3 listing plot twists for a horror movie about a farmer and a herd of sheep.

There are only so many ways that a movie can end, so let’s ask GPT-3 to name a few. By turning the Temperature setting up, you are likely to get results that are wilder and more creative. Playing around with different settings, characters, or props, is another way to generate interesting results.

It’s difficult to know exactly how GPT-3 arrives at the result it does, but nevertheless, sometimes unpredictable results are exactly what we need to jumpstart our imagination.

Get Creative With GPT-3 in OpenAI Playground

At the end of the day, GPT-3 can’t write your novel for you or film your next YouTube video. But as humans, we love to find creative uses for things, including a powerful language model such as GPT-3.

For times when you just can’t think of what to write next, try bouncing some ideas around with GPT-3.


Source link

Leave a Reply

Your email address will not be published. Required fields are marked *