AI Art: When Getting Worse is the New Perfection

When I first read The Verge’s headline – “AI image generators are getting better by getting worse” – my mind, as Cortex News editor, was instantly hooked. How can something improve by deteriorating? I dug deeper and, believe me, the explanation not only makes sense, but reveals an exciting phase in the evolution of generative artificial intelligence. It’s a perspective that, in my view, redefines the very concept of “improvement” in a creative context and gave me plenty of food for thought.

The Paradox of “Worse” Unveiled

My initial interpretation quickly collided with reality: “getting worse” is by no means a technical regression, but rather an intentional expansion of model capabilities. In practice, AI generators are no longer aiming solely to replicate reality with photographic precision, nor to produce exclusively flawless outputs. Instead, they are learning to generate more varied images, more “human” in their flaws, and less homogeneous. This is an evolution that, in my opinion, leads us to consider AI not just as a reproduction tool, but as a genuine vehicle for creative exploration, even in its most abstract or surreal forms. It’s as if an artist, having mastered classical technique, turned to expressionism: they haven’t “gotten worse,” they’ve simply expanded their language.

From Homogenized Perfection to Creative Diversity

For years, the stated goal of many developers and the primary user demand was to make AI images indistinguishable from real ones. We were obsessed with the ability to generate perfect hands or symmetrical faces. But, in my opinion, this pursuit of sheer perfection also limited versatility. Now, I see a significant shift in direction. The newest models, while retaining the ability to generate “perfect” results, are increasingly trained to respond to prompts that demand unique styles, intentional imperfections, or even “errors” that we would have once considered failures. This allows artists and designers to push further, yielding outputs that don’t just emulate reality, but reinterpret it in unexpected and stimulating ways. It’s a bold step that embraces the complexity and subjectivity of art, unlocking unthinkable scenarios.

The Impact on the Creative Process and the Future of AI

This trend, in my view, is fundamental to the future of human-machine interaction in the creative sphere. It is no longer about passively delegating creation to AI, but about actively collaborating, leveraging its capacity to generate a myriad of variants—including those that might elude human imagination. “Deterioration” thus becomes a tool of serendipity, a way to discover new artistic directions that the pursuit of mere “perfection” would never have allowed. Personally, I find that this evolution further democratizes access to advanced creativity, allowing even those without technical skills to explore complex and stimulating visual worlds. The direction is clear: AI is not here to replace, but to amplify and transform our creative potential.

Ultimately, in my opinion, the idea that AI image generators are “getting better by getting worse” is one of the most fascinating and promising insights of recent times. It pushes us to reconsider what “quality” truly means in AI-generated art and to recognize the intrinsic value of diversity and even imperfection. It is a sign that AI is maturing, transitioning from a mere tool of imitation into a true creative partner, capable of surprising and inspiring us. But this raises a crucial question, one I often ask myself: how do you view this ‘deterioration’ of AI generators? Do you believe it opens new avenues for creativity, or does it worry you that average quality might decline? I am extremely curious to read your thoughts on this.