Thread by @zemotion

Sat with Ted Chiang at @NeurIPSConf lunch ytd & someone in AI brought up democratizing art-making.

It was a candid chat so I was like, hey, can I just say something?—I really hate the word "democratize" used this way. Just say "automating art-making." It’s more accurate. 1/

Ted put it better than me and asked, "Do you say we should democratize marathon running? Why not?

"If you want more people to run, you don’t give them mopeds." 2/

We had an interesting conversation about the impact of word choice and then went to his afternoon talk, which was a full house!

Took notes on some points I really liked: 3/

"When you write a story, you are consciously or unconsciously making a choice about every word you type. To oversimplify—we can imagine a 10k word story is something like 10k choices. 4/

"When you give generative AI a prompt, you're making very little choices. If you use a 100-word prompt, you're making 100 choices.

If AI generates a 10k word story based on the prompt, it has to fill in for all the choices that you are not making." 5/

"Art requires making choices at every scale.

Believing that inspiration outweighs everything else—that's someone unfamiliar with the medium." 6/

While I don’t think effort alone equates art, the way he brought up the choices we constantly have to agonize over as artists made me feel like our hard work was recognized somehow, & it really moved me. 7/

"Art comes at great cost.

If you want more art in the world, ask your government to fund more art schools, grants, maybe universal healthcare." 8/

Later at the fireside chat:

Ted: What is the purpose of people trying to build gen AI?

@YejinChoinka: It's like a leaderboard, and people want to climb leaderboards! 😅 9/

His closing message to generative AI researchers:

"How is what you're doing different from geologists working in the oil industry? Are you using your knowledge of geology to make the world a better place?"