Specific Arguments against open source LLMs? — LessWrong