Do you think western models would be sufficiently censor-y about the topics in the curriculum? In my experience they kind of try and be honest overall but havent tried super hard with different topics.
Do you think western models would be sufficiently censor-y about the topics in the curriculum?
That’s the thing we want to find out, instead of guessing.
For starters, Grok has a few “truths” hardcoded in its prompt (probably nothing from the curriculum).
Do you think western models would be sufficiently censor-y about the topics in the curriculum? In my experience they kind of try and be honest overall but havent tried super hard with different topics.
That’s the thing we want to find out, instead of guessing.
For starters, Grok has a few “truths” hardcoded in its prompt (probably nothing from the curriculum).