When machines learn to design themselves

Generative AI is changing how we think about creativity. Where does human agency fit in the loop? The first generation of AI design tools were assistive — they suggested, they optimized, they automated the execution of human direction. The emerging generation is different. It proposes, it iterates, it operates with degrees of autonomy that challenge our understanding of creative authorship.

This is not simply a technological shift; it is a conceptual one. We are being forced to reconsider what we mean by design, by creativity, by the human element in processes of making. The questions are philosophical as much as practical.

The Prompt as Design

In the new paradigm, the traditional activities of design — sketching, modeling, refining — are increasingly delegated to systems. The human role shifts upstream, to the formulation of intention, the specification of constraints, the curation of output. We become prompters, curators, art directors rather than makers.

Some lament this as deskilling, a loss of the craft knowledge that traditionally defined design practice. Others celebrate it as liberation from tedious execution, an expansion of creative possibility for those without technical training. Both perspectives capture part of the truth.

The deeper question is about agency. When a system generates hundreds of variations on a theme, and the human selects among them, where exactly does creativity reside? Is it in the generation or the selection? In the system's capacity to surprise or the human's capacity to recognize value?

"The future of design may not be about making things but about making decisions — about developing the judgment to recognize good outcomes among infinite possibilities."