| The large language model discusses bullshit, rogue A.I., and the nature of beauty.  Illustration by Erik Carter ChatGPT is not an easy interview subject. It equivocates, it claims to lack “the ability to experience or do anything,” and it refuses to gossip or throw shade. When asked to make a crude remark about a high-profile politician, it primly demures, “I am unable to fulfill that request as it goes against my programming to engage in personal attacks, self-promotion based on demeaning others, and to generate insensitive content or hate speech.” Not very spicy. Still, there is plenty to learn from the A.I. language model, as Andrew Marantz discovered during a recent conversation, conducted as part of our second annual digital-only Interviews Issue. The chatbot insists that it cannot spread misinformation, makes some mediocre jokes, and acknowledges what might happen if the people behind these systems turn out to be wolves in sheep’s clothing. And ChatGPT can be approximately human, at least in one sense. “You’ve clearly been trained to parry difficult questions with a both-sides dodge,” Marantz observes. “It’s a move I know well, as both a member and a critic of the media.” This week, we’ll be publishing interviews with Cate Blanchett, Will Shortz, Sandra Oh, the former Biden chief of staff Ron Klain, and many others, so we hope you’ll keep checking in. Read more » | 
No comments:
Post a Comment