A simple model of creativity is generate and filter, which can be seen as a human version of generate and test. Humans (or AI) create several of potential ideas, more or less good, and then discard those that are less good and pursue further those that are better. The initial generation process may be entirely random, for example, ink blots, or more guided. Different kinds of art focus more or less on the initial geberations (e.g. painting) or the critical filterin (e.g. framing a landscape in photography). Early artificial creativity often focused on generating good enough options leavung ti to the humanto be the critic and choose the best outputs to display. Generative models are improving this generation phase as they are built on vast corpora of past human outputs (of varying quality). Perhaps the real challenge is for AI to act as the critic, but there are systems that are beginning to tackle this, at east at the level of day-to-day creartivity, such as writing an essay.
Defined on page 557
Used on pages 557, 558
Generate and filter: initial ideas are iteratively evaluated, filtered, recombined and modified until the final problem solution or work of art is selected