Long before we lived in a world where you could look at Vincent Van Gogh's The Starry Night with a single tap on a screen, creators and their work were few and numbered. In fact, there were many artists and thinkers whose work took the world by storm only after they passed away. Think Emily Dickinson, Franz Kafka, Galileo Galilei, Edgar Allan Poe, and so many more revolutionary creators. Fast forward to a few hundred years after that time - not only does the audience have easy access to artists and their work, but creators also have a multitude of avenues to showcase their work when they please.
Now, imagine whenever you look at the work of contemporary writers, marketers, designers, painters, thinkers, and other artists and think of fallacies like 'creative people are born with the talent' and 'I don't have what it takes to be an original artist', you could instead have the power to tap into your creative neurons with technology serving as a helping hand. That's precisely what Generative AI is trying to be at its core.
Almost every time you discuss Generative AI - or AI and robotics in general - funnily enough, you start arguing whether it can replace human jobs. Up until recently, machines were assigned heavy cognitive analysis work. They can discover relationships in data and suggest which Netflix movie you'd enjoy watching next, detect faces at the security kiosk of an airport, and even predict how much traffic your route has. They, however, couldn't create original work yet.
So before we begin to answer if Generative AI can replace human creativity, let's understand a bit more about what it is and how it's different from traditional or Analytical AI. To put things in perspective, we asked a Generative AI tool to help. Chatsonic, an alternative to the wildly popular ChatGPT, was put to the test. We asked Chatsonic a simple question.
Write a paragraph comparing analytical AI with generative AI using a real-world example.
This is the answer we got.
We then asked Chatsonic to make the answer’s tone a bit more conversational. Talk to us like a travel guide, we said.
This is the updated answer we got.
Here were our first thoughts on this interaction.
But are they immensely creative? Is this exactly how a content writer or data scientist would explain the question under consideration? That’s up for debate because creativity is subjective.
Instagram turned us into bloggers. TikTok turned us into performers and entertainers. iPhones turned us into photographers. In fact, lately, people have been able to turn social media platforms into multiple sources of income. Clearly, technological advancements have helped us broaden our creativity, changed how we create content, and altered how we perceive traditional forms of work or earn a livelihood.
Now, we’re turning the dial on how we create original content with Generative AI.
Built on top of large language models, input-to-text is currently the most advanced application of Generative AI. Given a well-structured and thoughtful prompt, these models are reasonably good at writing short-form content and even descriptive paragraphs. A little fine-tuning to the generated content is all you need to get it ready.
Natural language, however, takes a lot of work to nail. Given the intricacies of phonetics, different interpretations of emotions, and more, only time will tell how advanced these text models can get. Plus, it's also hard for a machine to figure out if its responses are morally and culturally appropriate or not. Take, for example, the many instances of ChatGPT spitting out racist, sexist, insensitive data on multiple occasions. Here's one example:
At the end of the day, organic flow in local languages, well-fitted emotions (sarcasm, irony, happiness, quirkiness, sadness, melancholia, etc.), and overall quality are crucial for original content to not just be unique but also engaging and impactful.
And so Generative AI-powered images had their own moment in the Sun recently. These image models are pretty wow-worthy thanks to the many aesthetic styles they can generate and how sensitive the model is to prompts. So why not check it out?
We asked Midjourney, a Generative AI tool that generates images using textual descriptions as prompts.
First, we fed a simple prompt.
cat playing with a yarn in the meadows while there's a storm incoming
Here’s what we got.
The model aced the colors and accents one would usually expect to think of when there’s a storm in question. The moodiness on the cat’s face also reflects the weather. Good job, we’d say.
Then, we tweaked the prompt a bit.
a sweet-looking cat playing with yarn in the meadows while there's a storm incoming during the day
While this might not look like the sweetest cat out there, notice the slight changes in its facial expressions and body language. The cat's brows look more relaxed in this version than in the first one, and its body also looks a bit less tense. Plus, Midjourney was sensitive to the 'during the day' addition to the prompt and balanced out the light from the Sun with the gloomy greys of the storm.
We thought we’d take it a bit further. Mission: Get the cat to throw us a little smile.
Here’s the prompt we fed it.
a smiling cat playing with yellow-colored yarn in the meadows while there's a storm incoming during the day. There are butterflies and cosmos flowers near the cat
Lo and behold! We’re finally graced with a beautiful smile.
However, notice how differently the model interpreted the 'yellow colored yarn' part of the prompt. Without explicitly mentioning 'whimsical' or 'magical' words in the prompt, the tool somehow gave it that sort of a spin. We're still trying to understand why.
Last but not least, text-to-video generative AI models are quickly coming into their own. Imagine the potential for text-based input to generate videos. They could be used to create videos from text-based scripts, allowing faster and more efficient video production. These models could also generate videos from text-based descriptions of a scene or environment - for example, in virtual reality applications - allowing for more immersive experiences.
Another potential application of these models is in the creation of video games. Text-to-video models could generate game environments and levels from text-based descriptions, creating more complex and detailed game worlds and realistic visuals.
Finally, text-to-video generative AI models could also be used to create 3D models for physical product design, allowing for a faster and more efficient turnaround.
When it comes to creating original content, creative block is every creator's Achilles heel. Given the saturation of the creative field from marketing campaigns - by brands big and small - to published books to films and TV shows, creative blocks are inevitable. They are also frustrating and debilitating to the creative process.
Generative AI can provide a creative boost by offering a set of parameters and goals to help guide the creative process. This can give creators a sense of direction and help them see their work differently. Or reduce the time it takes to develop an idea. Or even generate unique ideas and content that can inspire and motivate.
The inflection point of Generative AI has created a gaping opening for tremendous innovation and reinvention. What seems like wishful thinking right now could snowball into reality. That also means the technology has a long way to go before we can even consider it a 360° replacement for creativity. Till then, however, Generative AI can be a tremendous accelerator for creative thinking.
Type.ai was used to clear creative blocks when writing this blog. The illustration for this blog was generated using Midjourney.