ChatGPT launched this month, and with it a whole heap of creative, funny, fascinating and frankly hilarious use cases.
The team at OpenAI continues to wow the world with its incredible deep learning language model.
But before we get into the details, here’s all the background you need to know.
What is ChatGPT and how is it different to GPT-3?
ChatGPT is a language model comprised of algorithms that analyze a truly massive body of text – scraped from the internet. Because of this, it can respond to questions or utterances on an extremely broad range of topics.
The creators OpenAI last year launched GPT-3 – a similarly huge and impressive language model. This is the model we used in our digital human, Sophie, to open up her language capabilities to speak about virtually anything at will.
ChatGPT, meanwhile, is that same model specifically designed to be a chatbot. It is reportedly better at following the context of a conversation, and giving conversational replies.
Could we have gotten ChatGPT to write this article?
Well, no. The data that forms the knowledge of ChatGPT is essentially the internet. OpenAI is clear to point out when you ask ChatGPT to browse the internet for you that the tool is not for that purpose.
Can it write a unique article for you on a topic? Yes. Can it find you the funniest topics being discussed right now? Luckily for this writer, the answer is no.
One day these tools might be great for pulling feeds of live data. But for now, we’re stuck with playful and creative anecdotes shared online by real people experimenting with ChatGPT.
Fortunately, that’s more than good enough. Here are six examples of ChatGPT conversations that are sure to brighten your day.
1. Writing sub-par (but proficient) rap
If you’ve ever seen a child try to rap, you’ll likely recognize the bars spit by ChatGPT when LinkedIn user Michal Klar asked the algorithm to provide lyrics for a song on the future of meat.
The hilarious AI submission includes two verses, a bridge, many a dubious rhyme and a chorus that refrains “the future of meat, the future of meat. What’s it gonna be like, can’t nobody beat”.
Jay-Z it is not. That being said, with a lot of refinement and a little more training, could AI write the next chart-topping hit? Potentially.
If nothing else, it shows how it’s now possible for generative AI to script a huge range of content, including creative pursuits like lyricism. Whether you’d wish to use it or not is another matter.
2. Coming up with creative briefs for AI generated art
AI art generators like DALL-E and Midjourney took the internet by storm this year. We covered how Heinz used AI art in our post on awesome uses of AI in marketing. But while these AI tools are amazing at taking an idea and bringing it to artistic life, they still require one thing – imagination. The user is required to come up with the concept, and generative AI does the rest.
Not so when you take into account ChatGPT.
Twitter user Guy Parsons told ChatGPT he was trying to come up with some “interesting, fantastical ways of decorating a living room”. ChatGPT generously did the creative heavy-lifting – including individual concepts for fairytale, enchanted forest and futuristic space rooms.
The user then copy/pasted these ideas into Midjourney. And voila – AI creativity with minimal human input.
3. Providing recipes without the "saga"
If you’ve ever tried to source a recipe online, only to find yourself reading someone’s SEO-friendly life story preceding the useful part, you’re not alone.
Twitter user Steph Smith decided to skip the saga and simply ask ChatGPT for a recipe for rosemary chicken. And the results were seemingly spot on.
Not only did the model come up with the recipe, it didn’t come with the baggage, the superfluous story and the ads that often obscure you from getting what you actually want. In short, it was more effective than an internet search.
Which now begs the question of how ingrained in our lives generative AI technologies can become when they serve such a function. On a near limitless number of topics, ChatGPT and similar tools could provide quicker and more direct answers than your search engine.
At very least, it’s food for thought. And food that won’t take you an age to find the recipe for, either.
4. Writing on-brand, legally sound conversation
File this one under “fascinating”. Because a difficulty so far (and we stress this is very early days) in AI-generated content is the risk of the content being off-brief, off-brand or otherwise legally risky.
We’ve been experimenting with it for some time, and have found the need to put guardrails in place to stop digital humans from exploring certain topics. GPT-3 does, after all, get all its content from the internet.
In a similar way to our guardrails, ChatGPT appears to make it less risky to rely on these AI tools to create customer-facing content.
Our partners over at Element X have been experimenting with this very concept, asking ChatGPT about the Consumer Guarantee Act in New Zealand and how this might relate to, say, a television purchase.
As Element X Co-founder Ming Cheuk notes, OpenAI has trained ChatGPT “to avoid giving advice when not appropriate (e.g. if the answer is sensitive, has legal implications or there is no definitive answer).”
It will certainly give brands a lot to consider, and a lot more confidence that they can rely on AI-generated conversations without it crossing any legal boundaries.
You can see Ming’s full video here.
5. Coming up with brand names
For anyone who’s agonized over the arduous and indelible process of naming a new product, or even your entire brand, you’re somewhat in luck. ChatGPT is able to come up with list after list of names, should you wish to ask it.
As quirky copywriting master Dave Harland shares on LinkedIn, you can even refine your line of querying to the nth degree, asking the algorithm to make the name less boring, more purpose-driven or even rhyme – with admittedly mixed results. Other examples, including our own experimentation, shows that ChatGPT is in fact able to rhyme more often than not.
Interestingly, the conversation also brought up another quirk of ChatGPT.
Yes, one of the many benefits of AI chatbots, virtual assistants and digital humans is that they’re always polite, always available and always responsive. But in this case, it's to a fault.
ChatGPT appears to refuse to let Dave get the last word when saying goodbye. Well, sometimes it’s really hard to let someone go, we guess.
6. Driving digital human conversation
As AI conversation tools become more human, it begs the question what generative AI language models can do when plugged into a digital human platform.
We happen to have one on hand 😉
You can chat to our digital human, Sophie, who uses OpenAI’s model to have open-ended, virtually endless conversations. Tell her about your day, ask her questions, share your thoughts or feelings, discuss your favorite hobbies or foods – she can talk about almost anything.
And if she happens to come up with anything similarly whimsical, creative or fun, we’d love for you to tell us on our socials.