CMV: art is a problem that generative AI will probably solve
Just to preface, I’m a writer and a visual artist, so believe me when I tell you I take no pleasure in this position. Well, maybe a little bit of spite, but the kind of spite where you’re stabbing your self and the other guy at the same time.
The way I see it, the task companies are trying to achieve with generative ai is the creation of a general purpose problem solving machine. As we’ve seen in the past few years, general models have gone from barely being able to consistently solve 2+2 to winning gold medals in the International Math Olympiad and solving or helping solve Erdos problems autonomously (regardless of potential contamination, this is not a feat that I would be able to pull off). It has progressed leaps and bounds with regards to coding, even if you don’t take Anthropic’s claims about mythos at face value, and moreover I don’t see any real reason why it won’t continue to get better at both of these tasks (an argument that I’m sure many of you will make).
For whatever reason, the art they produce has lagged behind. Part of this, I’m sure, is because companies have not been focusing on this as much as they have on replacing jobs that actually make people money. But I would like to believe that part of this is due to the genuine complexity of art and the human experience.
However, at the end of the day, ai is a general purpose problem solving machine, and art is a problem. The only conceivable difference I have been able to imagine in my brain, if you don’t assume magical things about human consciousness or art or whatever, between art and a very hard math problem is complexity. With the right amount of parameters and the right amount of compute, what exactly is stopping gpt 7 or Claude 8.3 (or whatever) from finally cracking the problem?