I agree with you here, Paul.

Once you go into defining values, beliefs, and desires, you realize just how problematic is the idea of AI alignment.

I was just stating the generally accepted description of the problem, but I think it's way more difficult than some make it seem - and they already say it's quite hard.