I completely agree: if the (hypothetical) perfect LLM wrote the perfect book/song/poem, why would I care?

Off the top of my head, if an LLM generated Lennon’s “Imagine”, Pink Floyd’s “Goodbye Blue Skies”, or Eminem’s “Kim”, why would anyone give a fuck? If it wrote about sorrow, fear, hope, anger, or a better tomorrow, how could it matter?

Even if it found the statistically perfect words to make you laugh, cry, or feel something in general, I don’t think it would matter. When I listen to Nirvana, The Doors, half my collection honestly, I think it is inherently about sharing a brief connection with the artist, taking a glimpse into their feelings, often rooted in a specific period in time.

Sorry if iam14andthisisdeep, I don’t think I am quite finding the right words myself. But I’ll fuck myself with razor blades before I ask a predictive text model to formulate it for me, because the whole point is to tell you how I feel.

I still feel this whole conclusion is akin to “we won’t need money in a post-AGI world”. An implied, unproven dream of AI being so good that X happens as a result.

If an author uses LLMs to write a book, I don’t give a fuck that they forget how to write on their own. What I do care about is that they will generate 100 terrible books in the time it takes a legitimate author to write a single one, consuming a 1000 times the resources to do so and drown out the legitimate author in the end, by sheer mass.