Let's assume for a moment that all design ideas, past and present, would be collected and digitally represented in a giant data bank. And that the digitised data would be retrievable and processed by artificial intelligence, or rather machine learning, a subfield of AI. That is, to use algorithms trained on data to produce adaptable models that can perform specific tasks such as sorting images or analysing big data that would result in sets of recommendations. For example, recommendations that would bridge the gap between architectural design, engineering, and construction enabling architects to work more efficiently by automating all repetitive, mundane and time-consuming operations and thereby free up time and resources to reflect on practice as well as further experimentation and speculation. Indeed, the design process is both iterative and reflective, that is, to reflect on one's actions so as to engage in a process of continuous learning (Schön 1983). But teaching computers to be creative, or rather learn from experience and adjust to new inputs, is inherently different from the way humans learn. Generative artificial intelligence, GAI, however, can help us generate new ideas and insights by analysing large amounts of data. So the current leading opinion is that GAI can guide, support and augment the creative process and help problem-solving. A fact that also underlies many of the most popular AI-assisted services – from Google to Netflix to Xbox’s Game Pass service. The number of AI-generated recommendations, however, can be overwhelming to the point of distraction and so it is important for the complex machine learning models to make specific and customised as well as contextualised recommendations. Machine learning, then, through a recommendation system, has the potential to become a powerful design tool. A tool not to replace a designer, but to assist and therefore enhance their work.
