Exploring Creativity Through the Lens of AI Models

Food for Thought

I am fascinated by artificial intelligence because it allows me to understand how our brains work, which is, apparently, what initially motivated some scientists to replicate brain mechanisms in computers.

Have you ever wondered why some people are creative while others are not? Some people are highly knowledgeable, but they don’t create anything from what they’ve learned. Even when they do, it’s usually motivated by the need for survival, like getting college degrees and jobs. For me, learning anything is motivated by the prospect of creating something with it. After all, why buy tools and ingredients if you are not going to cook anything?

The difference between the “discriminative” and “generative” models in machine learning helps me understand where this difference comes from.

In the illustration above, imagine the blue dots as “chairs” and the green dots as “tables.” Each has certain “features” that make us want to label them as such, but they are not identical, so they end up scattered, but many of them are similar, so they form patterns.

The “discriminative” way of thinking simply draws a line between the dots to classify them into “chairs” and “tables,” which means it is not aware of the patterns the dot forms. It is only concerned about where it can draw the line. The “generative” way of thinking does not draw a line but tries to vaguely identify the shapes and centers.

If your brain is predominantly discriminative, you can pass exams for correctly classifying objects, but you won’t be able to build a chair because you have no understanding of its features.

As an example, being able to classify different types of music, no matter how many tracks you listen to on Spotify, won’t allow you to write a song. To be a songwriter, you must have a “generative” understanding of music.

Still, I’m not sure whether the desire to be creative drives some people to understand the world generatively, or they have a predisposition towards generative thinking, which, in turn, makes them want to create—a chicken-or-egg paradox.

FYI: ChatGPT told me my analogy here is valid.