- cross-posted to:
- technology@hexbear.net
- technology@midwest.social
- technology@lemmy.ml
- cross-posted to:
- technology@hexbear.net
- technology@midwest.social
- technology@lemmy.ml
cross-posted from: https://lemmy.ml/post/36807834
cross-posted from: https://lemmy.ml/post/36807834
Ok reading further, i think i agree with the author on the way creativity functions and perhaps misunderstand what they mean by democratizing creativity. That being said, theres some places i think the author is reaally overestimating what LLMs are capable of. E.g.
This is a pretty sweeping statement that i think is wrong. An LLM cannot visualize worlds that could be. Its a statistical model throwing together statistically linked output. The LLM (i hate that the author uses the word AI, its become a meaningless term) spits out the statistically linked tokens, it doesnt go in with an idea of a world or visualize a world that could be, it shoves together the tokens that best match its input.
Idk, maybe im way off base here
No I agree with your initial criticisms of the little essay.
Copying the most relevant bit from another comment I left here:
Here’s a couple examples of “creative use of an LLM” that I thought were fine and were truly creative:
Having some generative model generate video scenes following a theme that someone wanted and stitching them together into a music video that had a coherent theme for a song that he wrote. That took effort and the LLM couldn’t have made the whole music video.
Using one of those voice models trained on celebrities’ voices to make Taylor Swift sing “Get Low” by Lil John. Because it’s a funny idea someone wanted to see happen and they used the tools to make it happen. Now there’s major ethical problems with literally putting words into someone else’s mouth but that’s a different issue from whether it is a creative endeavor or not.