It would have been worth exploring your stance on AI for AI defense in general here, at least a little. Particularly as I guess from your writing that you are not a biologist, meaning your idea here is more interesting from a conceptual/strategic idea, rather than as a fully thought-through suggestion for countering this exact scenario. (At least this is a key reason why I upvoted your post, although I don’t find the actual suggestion promising. I agree with eniteris objections. But I want to see more of this and I support your thinking.)
Personally, I am currently very much for using AI for safety and targeted defense purposes. To be clear: NOT developing AI capability to do AI safety work, but to utilize existing models for aiding with defense and safety work. I see it as a no-brainer, really. (I also actually think that that working on improving public safety sets a good example of what we should use LLMs for, vs. AI slop and cooking recipes.)
This suggestion of yours would fit into this category, from my POV, but it seems you may disagree? What exactly IS your overall thinking on using AI for defense and/or AI safety?
It would have been worth exploring your stance on AI for AI defense in general here, at least a little. Particularly as I guess from your writing that you are not a biologist, meaning your idea here is more interesting from a conceptual/strategic idea, rather than as a fully thought-through suggestion for countering this exact scenario. (At least this is a key reason why I upvoted your post, although I don’t find the actual suggestion promising. I agree with eniteris objections. But I want to see more of this and I support your thinking.)
Personally, I am currently very much for using AI for safety and targeted defense purposes. To be clear: NOT developing AI capability to do AI safety work, but to utilize existing models for aiding with defense and safety work. I see it as a no-brainer, really. (I also actually think that that working on improving public safety sets a good example of what we should use LLMs for, vs. AI slop and cooking recipes.)
This suggestion of yours would fit into this category, from my POV, but it seems you may disagree? What exactly IS your overall thinking on using AI for defense and/or AI safety?