I love art; I think it’s one of those things that makes like more meaningful and valuable. I think everyone should do something artistic. I have several artistic hobbies (which I’m generally very shy about, but value very much), deliberately live in a neighborhood full of art galleries and live music, and so on.
And yet, it feels like AI is coming for art.
Artists are scared. Actors are coming out in support of the AI safety bill that just passed the California state legislature:
In the writing world, things are even more of a mess. In games, Sins of a solar empire’s highly anticipated sequel launched with AI character portraits (much worse than the originals) and people are mad. Fiction editors like Neil Clarke, whose work I really love, are very concerned about AI in writing, and at the same time are overwhelmed by low quality AI content:
On the other side, we have Runway inking a deal with Lionsgate to train on film data and Stability AI signing James Cameron as a board member (tweet). A bunch of visual artists whose work I enjoy have expressed similar sentiments:
What this actually means for longer form video is probably similar to this well-known video from Sora:
This is a pretty fun video (watch it on youtube), but it’s really not fully AI. They actually had to do a lot of editing work to get a cohesive, compelling product (source).
AI Content Detection
This is complex, and I’m no expert, but my opinion is this:
Automating AI detection, especially in text (the main medium I care about), is very, very hard: it falters on newer models, and has a really high false positive rate on human-written text. To the point I think it usually will do more harm than good. But I’m no fiction editor…
AI generated text (again, especially for fiction, but for long-form writing in general) is usually super obvious and bland. It’s not “good writing” for most purposes — although it’s great for helping polish up scientific papers!
What I think
I usually don’t like AI art. I definitely make exceptions for some AI artists (Ryan Murdock comes to mind) who I think both do genuinely cool stuff and often train their own models.
When I’m writing for fun (like these blog posts, I don’t use ChatGPT, perplexity, etc.; anything at all. And I think it’s a real problem the way it’s been used to spam Amazon and even indie publishers (see Clarkesworld again) that just makes life worse for everyone; the arts are valuable and improve all our lives.
Honestly, I don’t think AI is that much of a risk to the kinds of art I love. I think it’s usually super noticeable, and not in a good way (see Sins 2 example above, or Neil Clarke’s many posts about it).
I’m also not as bothered by AI’s inclusion in movies and TV, especially as it seems to be “ethically sourced data,” but then, I don’t really like TV/movies much, so I’m not really one to judge or speak.
What do you all think about AI and the arts? Please let me know: why I am wrong about the artistic merit of using generative AI, or about how I should worry way more about gen AI destroying the arts, or some other thing I didn’t think of.
Conan O'Brien makes a similar point in his podcast: https://youtu.be/iekKf5sLuEk?si=GgzKz9i-802_ug-1&t=452
ChatGPT seems to capture a lot of the details of human art and everything, but there's always this eerie "synthetic" feel to it that no one can seem to put their finger on.
Arts are not going anywhere -- I think GenAI will make art even more accessible. Now you don't even need to learn to draw or paint to make a very good picture, or a whacky video.