Is it only torture if the numbers turn into English words at the end, or is it still torture if they just stay numbers? Would it be immoral to crunch the numbers by hand? What if I swap out the letter-string map at the end, turning "bad" into "good", "ouch" into "yum", and so on?
The kinds of people who work on AI are, by and large, atheists.
Anthropic isn't *getting* the opinions of religious leaders. They are trying to *affect* the opinions of religious leaders.
Any pretensions to the former are in service of the latter.
I don't know if scott is an ai safetyist, but if so this is a pretty good example of trying to score points with an ingroup at the expense of a more important goal.