LLMs should have been a public good

I don’t necessarily object to swallowing up all the world’s text and information. There is clearly value in what becomes possible. I object to:

1. Doing so without anyone’s express consent or input
2. Using literally all of the world’s available compute (multiplied by 4 companies doing this all at once)
3. Not sharing any of the motherfucking profits

In a different world, ChatGPT could have been something like Wikipedia. The dataset and the infrastructure could have been sourced and grown gradually from consenting parties, and the benefits could have been shared among us all.

I asked Claude what it thinks of this post and it gave me an even-handed critique.