DOC • The creator and the machine

Generative creation and the role of abstraction

At first, the introduction of computers to the creative community faced resistance, especially because of their mechanical, mathematical, and multidisciplinary nature. The core of the debate was the comprehension that creating artifacts alongside computational systems involved ceding a part of the creative process — which before belonged entirely to humans— to the machine. Such a shift naturally challenged the conventional conception of ownership, as it raised an intuitive question:

Who should be considered the author of a given piece of intellectual production, art, or design when computers are involved? The human, the machine, or both?

With time, and as computers got more popular, such resistance diminished as creators embraced digital techniques, whether by choice or pressure of the market. In any case, an important remark for our discussion is the realization that sharing authorship with machines is not something new brought by AI, but rather something that has been unfolding throughout the past decades and that was more recently intensified. This is also why we are now seeing a re-edition of the ownership debate I discussed before.

Nonetheless, the possibility of pairing with computational intelligence has since the beginning motivated professionals to explore these machines as a fruitful creative medium.

For designers, for example, accustomed to apprehending current technologies and repurposing them to the tasks at hand, such exploration came with many intents. Some of these were to extrapolate the capabilities delimited by available proprietary software (think of Adobe, Macromedia, Sketch, Figma), to obtain novel and unpredictable aesthetics, to enable the work with parameterization and optimization, or for building artifacts that respond more autonomously in real-time. In recent years, many designers have been leveraging generative creation very literally, building or employing systems that can render flexible visual identities, parametric objects, responsive interfaces, and generative fonts.

In many ways, ceding a part of our creative processes to machines allowed us to design more efficiently, accurately, and with greater—or at least novel—expression.

We say something is generative when it is capable of producing an outcome or reproducing itself autonomously. Therefore, designing with generative creation implies intentionally employing an autonomous element that contributes to the achievement of a certain goal or to the synthesis of desired outcomes. Such autonomy can be granted in several ways: by letting systems make choices based on complex models, by relying on sufficiently smart Gen-AI agents, by designing with genetic algorithms, or simply by designing systems to respond to unexpected interactions.

This is why we say generative AI is generative. Because such agents are capable of autonomously generating outputs and synthesizing artifacts, regardless of how predictable the input is. Some even learn from previous interactions to respond in smarter ways and produce even better.

Formally, generative creation means employing systems or processes, which are put into execution with a certain degree of autonomy, contributing to or resulting in a complete work. The critical point is that computational intelligence is intentionally used as an active participant in the creative process and not only to support the decisions made by humans (Groß et al., 2018; Grünberger, 2019; Galanter, 2003).

In design, working with such computational autonomy promotes a fundamental change in the creative process as designers become no longer executors of tasks, but conductors. A role that Groß and his collaborators consider to be that of an “orchestrator of decision-making processes” in their book Generative Design (2018). Essentially, bringing generative agents to our work means giving up total control, which is now partially conducted by a form of computational intelligence we need to manage.

To illustrate this fundamental change, Groß proposed a model for the design process around generative creation characterized by an emphasis on abstraction. The main change, according to him, is not only that traditional craft recedes to the background while abstraction and information become the protagonists, but also that designers need to constantly reflect on how to translate their ideas into information that autonomous agents can “understand”.

Thus, the relevant question is no longer “How do I draw/sketch/paint?”, but “How do I abstract?”.

Groß’s original illustration for the model focused on generative design through coding, but I amended it to highlight the possible role of generative AI agents (in blue).