AI started impacting creative fields first. Surprisingly, these fields, especially the arts, were generally believed to be safe from AI intrusion, more so than other non-physical professions.
AI researchers were probably aware of this and dedicated their work to developing creative AI. After all, displaying creativity is a surefire way to impress. A synthetic intelligence that could code proficiently, devise algorithms, or outperform humans in solving mathematical problems wouldn't cause much surprise. For most people outside the AI field, this would merely be seen as a glorified calculator.
So, how to convince the world that AI is serious business? The answer lay in targeting the arts - widely considered the pinnacle of human intellect - and that's precisely what AI researchers did.
Google's Deep Dream, released in 2015, was the first AI that stunned people by producing psychedelic, dream-like images. Then, Google's Magenta arrived in 2016, using machine learning to compose classical music. In 2018, we saw the arrival of GPT, a machine learning model capable of producing human-like text.
But until GPT-3, none of these technologies seemed practical, and many people thought that the realm of human artistic creativity would remain in human hands. This assumption, however, was flawed. Like language, mathematics, and logic, the arts are abstract in nature, and abstraction is inherently a non-biological process.
Consider the Ebbinghaus illusion as an example. Two orange circles of the same size, when surrounded by differently-sized gray circles, appear to be different sizes. Yet, when we abstract from our biology and use a measuring system, we verify that they are, indeed, the same size, despite what our eyes tell us.
The introduction of DALL-E and GPT-3 caused quite a stir. Their successors, GPT-4 and Midjourney, took things to an entirely new level. Text-to-image synthesis became phenomenally accurate and artistic, sending shivers down the spines of artists everywhere.
This approach works so well because it mirrors the way humans create art. Humans observe the world around them and synthesize novelty by combining aspects of that world. Over time, this process of observation, abstraction, and synthesis led to inventions like the airplane.
The concept of a purely original idea is a fallacy. Every idea is a result of input from the external world. While it may seem as though creations and ideas spring purely from human minds, they can all be traced back to external inputs from long ago.
Yes, randomness, accidents, and our recognition of them can produce original ideas, but these would still be collaborations between humans and natural forces, not truly original human thought.
AI generates content much like humans do, by combining inputs. And AI can do this far better, thanks to its ability to handle a much larger amount of inputs.
So where does this leave artists, creatives, and thinkers? In a precarious position, indeed.
Even the current versions of AI can combine random inputs to generate images, articles, stories, mechanisms, and so on. Factor in randomness or 'errors', and you have a recipe for unlimited creation.
However, we're not interested in just any creation. We're focused on specific use cases.
Let's return to art. Suppose an AI has generated all conceivable artwork up to the present. This would result in a potentially infinite number of artworks, most of which would not meet human standards. But when a specific use case is applied - let's say, an art exhibition themed "Seasons" with a blend of old-world masters' style and modern aesthetics - the pool of AI-generated works is significantly narrowed.
The task for artists now is to choose the best artworks that fit their taste. The challenge isn't creating art - since it already exists - but selecting from the existing works to find those that both fulfill the needed purpose and match their aesthetic preference.
This process bears striking resemblance to traditional art creation. When producing a piece, an artist crafts many versions before deciding on the one that best communicates their intended message.
The key factor isn't the act of creation itself, but rather knowing when to cease, when the piece is complete.
In essence, the principal role of an artist, musician, storyteller, or creator is to serve as an agent of taste.
Consider a painter, whose work process, much like an AI diffusion model, involves an act of refinement. Each stroke isn't an addition to the canvas but a subtraction of possibilities. Life decisions follow a similar pattern. Choosing to become a medical doctor at age 18 and pursuing that path forecloses other potential life trajectories.
In the end, what sets a good artist apart from a mediocre one comes down to taste. All technical aspects can be learned given enough time and effort. Yes, some musicians can play faster than others, and some painters can create photorealistic images that many others can't. But music isn't a speed competition, and painting isn't photography.
An artist, regardless of their medium, learns theory and hones their technique to produce work that lives up to their taste. Taste, however, can't really be taught. It is the culmination of one's unique exposures and experiences, resulting in a distinct artistic aesthetic.
This aesthetic, or taste, has always been the most crucial facet of a creative individual. And in the age of AI, it will persist as the most important and relevant role for creatives.