As technology increases in its capacity, one of the arguments I’ve heard is that humans are able to be creative, but machines and computers are able to be efficient. Perhaps a computer program can crunch numbers in a spreadsheet faster than a human can add them in their head, but no computer can create a masterpiece of art.
However, I wonder if this argument for human uniqueness may be waning as technology continues to advance. For instance, consider this video (produced by ING Bank as a promotional piece for their innovative approach to technology and data):
In this case study, undisputedly, the creative genius at the center of the story is Rembrandt. The derivative work is the algorithmic analysis and duplication of something that looks (to some degree) like another Rembrandt painting.
However, there is a level of creative genius in conceptualizing and painting a new Rembrandt painting! (Surely we give this much credit to the original artist). This is especially the case because there was no subject to be painted - the portrait is, interestingly, a portrait of no one in particular.
Second, this increases the likelihood that an algorithm could one day be trained, not only to duplicate an existing set of artistic work with small modifications, but to create a piece of art that is more substantially new.
A parallel example would be AlphaGo. The first AlphaGo learned how to play Go at a superhuman level by studying an incredible number of Go games played at an expert level. Then, AlphaGo Zero learned how to play Go even better than AlphaGo by starting from scratch. They simply let one copy of Zero play another copy of Zero and the resulting approach to the game of Go was dominant over the original AlphaGo.
As the researchers for AlphaGo put it, “This technique is more powerful than previous versions of AlphaGo because it is no longer constrained by the limits of human knowledge. Instead, it is able to learn tabula rasa from the strongest player in the world: AlphaGo itself.”
This example makes me think that a new master artist AI program might be around the corner. The first version duplicates Rembrandt, just like the first AlphaGo duplicated Go players. But the second version may go beyond what Rembrandt was capable of doing. And the third version may go beyond what any human artist can achieve.
What do you think?
- Is it possible - or even likely - that computers will produce superhuman quality artwork in the next decade?
- What implications will this have for the social perception of AI? For our self-perception of human value and uniqueness?
- In what ways do these AI programs fit into the redemption of the world? Or work against God’s plan for humanity?