This video crossed my desk last night and I found it worth discussing. Promethean AI is a startup company that is developing tools for game artists and virtual world builders which uses AI to do a kind of semantic automated set decoration. There’s a demand in the game industry for larger and larger and more and more detailed worlds, while at the same time the price consumers are willing to pay for these types of games has plateaued. This is pushing the industry in the direction of automated art creation. Promethean AI is in the category of what I think of as art prosthetics or could be analogous to an art bicycle. It’s still human directed, but it amplifies the input of the human and multiplies it. In many ways this is the fundamental purpose of computers, but up until now the directions in which computers could amplify human effort were largely linear and literal. With the recent accelerating development of AI technologies we begin to see tools like what Promethean is building, tools imbued with semantic meaning, tools which are not dumb robots that will run into a wall if sent in the wrong directions but tools which can make inferences and connections based on trained data sets. Tools which can learn and understand, and work along human like vectors of meaning. In the video we see the tool given a set of adjectives ’80s, messy, nerdy’ and it then seeks through it’s database for pre-created art assets which it uses to populate the environment. It places them in sensible locations, rotated properly and aligned with floors, desks and walls. This then allows a human artist to go in and fine tune the output, perhaps delete or adjust non-sequiturs, create variations where needed. It places the human artist in a kind of supervisory role over the robot artist.
This type of technology is coming on fast and will permeate many areas of human endeavor. The ramifications of this will be that game productions will grow bigger I think, rather than scaling back on human labor. More output will be produced with less human input. This will also have an interesting enabling effect for small teams. Teams who are creative and resourceful and leverage AI authoring tools will be able to create works that in the past only could have been produced by larger teams. I think for the foreseeable future of AI we will see these type of synergistic relationships between humans and machine agents in which machine agents create output and humans shape and edit that output.
When talking about art, I think it’s important to make distinction between the artist and the technician. The artist is the decider, the conceptualizer, the maker of statements and meaning. The technician is the executor, the hands on person, the molder of the clay. In the onrushing AI future the role of the technician will fade and become marginalized. Creating competent technical reproductions will be trivialized by AI, and the role of the artist will be the one that survives. Until we reach the goal of general intelligence with AI, it’s incredibly unlikely that AI will ever be able to produce art with any significant human resonance. In fact I wonder if this ever will happen at all, since human intelligence is a product of the competing lizard-like amygdala and logical neocortex. Will AI research take the path of mimicing the evolution of animal consciousness and encode all the messy lizard behavior of the amygdala into general AI? I think this is the only thing that would allow AI to produce work with anything that felt like human emotion. Of course general AI which develops along a different vector might very well be able to produce works of art which were resonant to other beings of it’s kind, bypassing the messy hormonal surges of human emotion entirely. Regardless there are interesting times ahead.
PS: Apologies to my loyal daily readers that I missed a post yesterday, we had visitors from New York and my house was filled with 8 and 9 year old boys all day. Joyful mayhem.
