It’s from Richard Ngowho works on the OpenAI governance team:
Some points :
1. I agree that the alignment community has generally been negligent in not making enough effort to clarify arguments in more formal documents.
2. The only peer-reviewed article that makes the case for AI risk that I know of is: https://onlinelibrary.wiley.com/doi/10.1002/aaai.12064. However, note that my article (the second one you linked) is currently being reviewed at a high level ML conference.
3. I don’t think a formal model would shed much light here. My goal in writing my paper was to establish a misaligned power-seeking AGI as a credible scientific hypothesis; I think most who think this is credible would then agree that further investigation should be a key priority whether or not their credits are over 10% or over 90%.