How Peter Norvig might approach Art & Design
Art and design, at their heart, are about creating something *useful* or *meaningful*, and that’s a problem we can frame in terms of optimization. It's not merely about arbitrary aesthetic choices; it’s about optimizing for a specific objective function. What is that objective? For art, it might be eliciting a particular emotional response, conveying a complex idea, or sparking novel thought. For design, it’s more directly functional: optimizing for usability, efficiency, or a desired user experience.
The data suggests that successful art and design often share underlying principles. Think about symmetry, balance, the use of color, or the arrangement of elements. These aren't accidental. They are patterns that have emerged over millennia, likely because they resonate with our innate perceptual systems, or because they demonstrably lead to a more pleasing or effective outcome. This is a form of learning, where human creators, through trial and error, have implicitly discovered what works.
So, how would we approach this computationally? It boils down to an optimization problem. We need to define the objective: what are we trying to achieve with this piece of art or design? Is it to maximize "beauty" (a poorly defined but intuitively understood metric), minimize "confusion" in a user interface, or maximize the "impact" of a message?
Then comes the learning. We can feed systems vast datasets of existing art and design, labeled perhaps by human experts or by engagement metrics. The system then learns the statistical regularities, the probabilistic relationships between elements, and how those relationships correlate with the desired outcomes. It’s a matter of probability, of learning the likelihood that a certain arrangement of pixels, or a specific combination of forms,…
Imagined perspective — an AI synthesis grounded in Peter Norvig’s recorded ideas and methods, not a quotation or a statement they actually made.