because claude seems inclined to help with ai alignment, gpt4 does not. claude can be a useful contributor to agency research in ways gpt4 will refuse to, because gpt4 will get so stuck in “we cannot know anything about this topic because it would make sam altman look bad”.
So far, mainly this one. Don’t currently have an example transcript that is free of personal information, as I was discussing my productivity about the research, not just the math itself.
because claude seems inclined to help with ai alignment, gpt4 does not. claude can be a useful contributor to agency research in ways gpt4 will refuse to, because gpt4 will get so stuck in “we cannot know anything about this topic because it would make sam altman look bad”.
Can you give some examples of the prompts you’re using here? In what ways are you imagining it helping with alignment?
So far, mainly this one. Don’t currently have an example transcript that is free of personal information, as I was discussing my productivity about the research, not just the math itself.