Ugh, it's not about LLMs. Us "doomers" don't know if you can make an intelligent system with a sufficiently advanced LLM as the core component. We mostly don't

Ugh, it's not about LLMs. Us "doomers" don't know if you can make an intelligent system with a sufficiently advanced LLM as the core component. We mostly don't think an LLM that's not part of a bigger system with appropriate feedback loop and storage systems is likely to be an X-risk. It's about whenever we DO make true AGI making sure it actually doesn't end up with goals we don't really want.