Yes, and I was attempting to illustrate why this is a bad assumption. Yes, LLMs subject to unrealistic limitations are potentially easier to align, but that does not help, unfortunately.
I don’t see how you’ve shown it’s a bad assumption?
Yes, and I was attempting to illustrate why this is a bad assumption. Yes, LLMs subject to unrealistic limitations are potentially easier to align, but that does not help, unfortunately.
I don’t see how you’ve shown it’s a bad assumption?