Sunday, March 22, 2026

Romanticising creativity?

GenAI models, such as ChatGPT use algorithms trained on vast amounts of data from existing work, and then generating their own novel content (text or image). But are the models really creative? To judge if GenAI models are truly creative, it seems important to keep in mind that the models don't understand the meaning or context of their creations directly and only generate outputs algorithmically based on the data they've been fed. In contrast, humans understand the context and its nuances naturally and also show a greater ability to come up with abstract ideas with great variations between individuals due to their unique and detailed experiences.That said, GenAI can certainly augment human creativity. It can generate many ideas, concepts, or rough drafts, which a creative person can then refine and build upon. In this way, GenAI can act as a co-creator in the human creative process. But ultimately, does it all just come down to the fact that humans provide greater variations than GenAI due to their unique and relatively diverse (multi-modal) experiences of the world? Or, are people overly romanticising human creativity? Because isn't creativity essentially remixing parts and recombining elements of thought, or synthesising findings or creating metaphors? If so, the limits of creativity, it seems are the limits of human language. And GenAI is pretty good at that too, that is, they are making word-by-word predictions in the context of human prompts. Yet, GenAI doesn't have the same kind of complex directionality as humans with their memories, intuitions, dreams and wishes. In short, GenAI is not an individual. This fact, however, highlights the ethical, social and human costs of GenAI, harvesting human knowledge without respect for privacy and copyright boundaries: credit back to the humans!

No comments: