Raw LLM Responses

So, as someone who is primarily a programmer (legacy software), and also dabbles in 3d modelling and drawing, but doesn't see a problem with GenAi. 1. The issue is the speed at which neural networks are able to process data into statistical features? Also, "stealing" is a deliberate wording. Are you stealing when you view an image on the web or save it on your computer? Because in terms of IP/Copyright infringement, the illegality is redistribution/republishing. The training dataset is not available or part of the final weights of a neural network during inference/generation, so collaging by it's literal definition is not possible. 1-a. Payment. How would you propose any money would be distributed? Datasets for modern models like StableDiffusion or ChatGpt/Gemini image generation require billions of images. With a 10% royalty distributed to the artists/photographers, it would take tens of billions in revenue just for each image in the dataset to generate $1 in royalties. 1-b. Legality. It has been ruled in some places such as Japan that training AI on copyrighted works falls under fair use as it doesn't redistribute or republish the original work. 2. I don't disagree with the general sentiment. 3. Don't share the tech doomerism viewpoint, but overall don't disagree with the rest of the general philosophical sentiment.