3 ms·

It's usually pretty easy to get gpt to moralize about why "i cannot help you make that shell script because it requires the usage of "kill"", but i thought it w

It's usually pretty easy to get gpt to moralize about why "i cannot help you make that shell script because it requires the usage of "kill"", but i thought it was very weird that while it's usually very aware of biases and possible sexism/etc, I could not get it to understand or give any possibility to why, maybe, Anthropic chose Claude as the name for their AI instead of Shannon because of mild sexism (Shannon is a girls name after all) We don't call Turning machines Alan machines. and Shannon sounds better, at least to me This is half tongue in cheek, I don't think that was the reason for the most part (maybe just a slight bit), but it's strange that while gpt will often talk about possible biases, it can't see the possibility of one in play here. Maybe because the topic is "AI"? Or can RLHF actually somehow prevent it from recognizing what might be actual sexisms/etc when it's done by a third party you're talking about, as opposed to back and forth with a chatbot?

revelio 3y ago
They chose Claude because if you're going for a friendly human sounding personality, a first name (which is how people address each other) sounds much better. How many people refer to each other by their last names in normal casual conversation, outside of maybe the military? Even ChatGPT in its heavily neutered state isn't silly enough to claim the Anthropic devs are trying to subconsciously oppress women via name choice. This constant witchhunt for sexism and racism is really tiring, btw. It's long ago blown past boy-crying-wolf territory. You should know that many of us now assume that any claim of sexism or racism is automatically false, just another attempt to dig up virtue truffles from already wrecked land. It makes the accuser look bad and worth avoiding, whilst having no effect on the reputation of the supposed guilty party.

revelio 3y ago

They chose Claude because if you're going for a friendly human sounding personality, a first name (which is how people address each other) sounds much better. How many people refer to each other by their last names in normal casual conversation, outside of maybe the military? Even ChatGPT in its heavily neutered state isn't silly enough to claim the Anthropic devs are trying to subconsciously oppress women via name choice. This constant witchhunt for sexism and racism is really tiring, btw. It's long ago blown past boy-crying-wolf territory. You should know that many of us now assume that any claim of sexism or racism is automatically false, just another attempt to dig up virtue truffles from already wrecked land. It makes the accuser look bad and worth avoiding, whilst having no effect on the reputation of the supposed guilty party.