1. Is everyone using LLMs to solve problems? No, they aren't. Not even close. 2. "...in a few years..." No, this has already happened. LLMs are producing outpu

1. Is everyone using LLMs to solve problems? No, they aren't. Not even close. 2. "...in a few years..." No, this has already happened. LLMs are producing output that is then posted on the Internet which thence finds its way to other LLMs (and back to itself). Much discussion about this has occurred. 3. "In short, how can ... [very long sentence]* the long term?" It can't other than by human editing. For the LLM model, curation of content is mandatory. Limit the LLM corpus to scholarly and scientific works and you'll get an LLM that is "smarter". Limit the corpus to bona fide literary masterpieces and you'll get an LLM that is a better poet. Suck up everything on the Innerwebs and you'll have a GIGO (Garbage In, Garbage Out) situation. The LLM model is too limited to produce intelligence. LLMs predict the next word and build sentences, paragraphs and documents. When trained on a readable corpus an LLM will usually write readable output.