AI & Creative Industries: Panic or Celebrate? [Episode 1/4]

Is creativity compatible with AI? It is impossible to make a detour on the question as the subject AI is omnipresent. 1st episode of a series of 4 analysis articles.

Point de vue

AI & Creative Industries: Panic or Celebrate? [Episode 1/4]

This is the subject of the moment: for 1 year or 2, it has been impossible to go a week without a summit, a conference or a webinar on AI and its consequences on industries with a strong creative component.

For the media, for the music industry, entertainment, for design and architecture, but also for the luxury industry or marketing and advertising.

In this kind of confusing debate where you can read everything and the opposite, and as a creativity speaker and an innovation speaker, I like to turn to science, in particular to cognitive psychology, neuroscience, and the sociology of technology. In order to get out of the passion dimension and to confront subjective conjectures, concrete and validated facts.

So here is an overview in 4 successive “episodes” of what science tells us about the AI x Human match on the subject of creativity.

AI is a great thing for creatives ” - heard in agency

First the consensus: AI models have remarkable performances. Nobody will deny it. They can generate creations so convincing that it is sometimes difficult to distinguish whether they are of human or algorithmic origin (DeepArt for the image, AIVA or Suno for the music, ChatGPT for the writing, etc.).

The fields of application are very varied and all creative sectors are impacted. With real, proven and measured benefit: advertising, product design, product design, architectural prototyping, musical composition, film creation, etc. benefit from an undeniable increase in productivity.

AI can speed up the creative process by quickly generating prototypes, offering suggestions based on data trends, and allowing customization at scale.

We are talking about productivity. Of efficiency. Of speed. But when it comes to creativity, that's not so much the subject.

Going back to the very definition of “creativity”

Research on human creativity shows that it is the result of a complex process combining intuition, emotion and reasoning. It involves cognition where unconscious and emotional elements come into play.

Creativity is therefore not just a simple accumulation of ideas for which we could confront a process, and therefore talk about productivity.

When science looks at the subject of creativity, there are several criteria for examining a “creative object.”

Creativity as a combination of novelty and value.

Scientifically, an idea or production is often considered “creative” if it is both new (original) and perceived as having value (aesthetic, functional, etc.)

So, is it a conscious phenomenon, linked for example to intuition (specific to humans) or can it be formalized by algorithms capable of recombining and innovating based on existing data?

Semantic comprehension vs. statistical recombination.

One of the major debates revolves around the real ability of AI to understand meaning or to be intuitive. Most current systems (deep neural networks) are based on statistical correlations, without real “understanding” or lived experience. Basically, it's statistics.

The question of intention.

According to researchers, human creativity often involves a conscious approach, a goal (or questioning). According to the psychologist Mihaly Csikszentmihalyi, author of the theory of Flow, creativity is a response to challenges that involves the profound capacities of the individual, especially in contexts of specific constraints.

AI, on the other hand, has no intention of its own. Scientists are therefore debating the difference between “emulating” a creative process (in the sense of simulation, in this case for AI with statistics) and “being” creative. We're almost getting into philosophy. To be or to simulate, that is the question.

Bias and standardization.

Finally, we know that AI models are trained on large databases. Some researchers therefore fear that, in the absence of data diversity or training methods, AI may reinforce aesthetic stereotypes and produce styles that are too similar (“average” effect). Basically, she would have a hard time getting out of the box, thinking “out of the box.”

But opinions are not unanimous on this subject...

Episode 2: Average effect and risk of homogeneity, coming soon