What is Generative AI And Can It Create Original Content?

Understanding How It Works and Evaluate its Creativity

From OpenAI’s ChatGPT to Google and Facebook throwing their hats into the ring, the tech giants are all over it. Photoshop, Roblox, Amazon… everybody jumps on the bandwagon with their new Generative AI features and people started creating content in ways we could only dream of a decade ago.

But while Generative AI is making its way into all industries, the question becomes: Can it really create something original?

In this post, I’ll try to cut through the hype and explore what generative AI is, how it works, and whether its creations can be considered original.

Understanding Generative AI

Generative AI refers to a type of artificial intelligence that is capable of generating content such as images, text, or music. The term “generative” emphasizes the ability to create new content by learning and utilizing patterns and relationships from training data.

Hence the term “AI model”: because this process builds a mathematical model that represents the patterns and relationships found in the training data.

How does an AI find patterns and relationships?

Actually, it’s through trial and error. Training here means that, over numerous iterations, the model tries to come up with a connection between input and output.

To visualize this, imagine an AI model being trained with images of a bird, a dog, and a cat.

At first, the model might not be able to accurately distinguish between them. However, as it processes the images, it adjusts its network of mathematical parameters whenever it makes a mistake in categorization. Gradually, it learns the distinct features and attributes associated with birds, dogs, and cats, and can accurately recognize similar images.

Similarly, this concept applies to other tasks you want an AI model to accomplish. For example: In the case of ChatGPT, the training process involves presenting the model with fragments of text and allowing it to predict or guess how the text should be completed. Again, the model adjusts its mathematical parameters to improve its predictions based on the examples in the training data. Over time, it learns to recognize patterns and structures in language and is able to produce coherent and contextual text completions.

Common Misconceptions

It’s crucial to understand that an AI model exists in the form of a set of mathematical parameters (that can be stored in a file for later use or deployment). It is not a machine crawling the internet and then synthesizing stuff from the data it finds. It is a mathematical representation of a tiny part of the world, that part which it could glimpse at through its training data (for example, written language or images).

Using Generative AI

Once an AI model finishes training, we can interact with it. Depending on what the model has been trained for, it will use its structural knowledge of its favorite part of the world (aka mathematical representation of structure found in its training data) to interpret our input.

It basically works like this:

For example, ChatGPT would try to find a meaningful completion for the text you give it because it has been trained to complete texts. It has no other objective (that’s why it cannot care for facts, it was not built to do that). Midjourney, on the other hand, would try to find combinations of pixels that fit your text prompt to create an image from that. It does not care for the content (in fact, that’s a huge problem when it comes to NSFW imagery, which Midjourney recently tackled in a pretty effective way).

Can Generative AI Create Original Content?

First of all, when people talk about “originality” they’re using a concept that is closely connected to creativity.

Unfortunately, this is where it gets a little prickly.

Not because we cannot tell whether Generative AI is creative or not, but because people have so many definitions of what creativity really is (or even: should be).

I wrote about some general aspects of the intersection of AI & creativity here:

In this post, however, I want to stick to a more technical view. Because since you know how Generative AI basically works, you have already formed an opinion on whether this seems to be creative behavior — for you.

And I would really love to hear about it in the comments!

But whether you find Generative AI capable of creative tasks or not, here are the bare facts of the matter:

• We have seen that every AI-generated content depends on the task the respective AI model is built to fulfill and on the training data that is used to create a mathematical representation of the world that helps it fulfill that task.
• This means: The properties and characteristics of the generated content, such as patterns, structures, and relationships, align with those found in the training data.

Now this might sound as if Generative AI cannot generate original content, because “alignment of patterns, structures and relationships with training data” does not sound very creative, right?

However, it’s important to note that an AI model can use its structural knowledge to produce novel combinations and variations of those patterns, structures, and relationships! That makes the generated content itself original and not a copy of any specific example from the training data.

Conclusion?

The way Generative AI works is very similar to the way an artist creates a new painting by combining different techniques and inspirations learned over time, or an author creates a new story based on dramaturgical devices, tropes, and rhetorics he read in his favorite books.

Some already call this artificial creativity, others remain skeptical.

What I know from my experience as a writer is that AI models can inspire human creativity in seemingly infinite ways, and that during the communication between humans and machines, creativity emerges from this new interactive system.

It seems to me that it is no longer exclusively the human part that is responsible for creativity; and, on the other hand, the machine relies on human input and guidance to yield astonishing results.

➡️ Follow me to stay up to date on “AI &Creativity”. If you want to support my work, become a Medium member using my referral link and get full access to all my articles (140+ and growing) and those of thousands of other writers. 🙏

➡️ If you like my content, why not leave a “clap” at the end of this article, so more people can see it?