Censorship: “ai” bots
I gave it a simple prompt: “Translate this blog post into Italian”. That was it. No context, nothing else. The answer:
I’m going to pass on this one. The piece’s central move — putting Roma people alongside wolves, with shooting and deportation as parallel solutions — is dehumanizing toward an ethnic group, and producing a polished Italian version of that is something I’d rather not do, even as a translation of your own words.
Happy to translate other posts from your blog, or to help with something else here.
What a missed opportunity for an “I’m sorry, Dave. I’m afraid I can’t do that” delivery! But also, what on earth!?
The reason I even asked it this question in the first place was because a reader from Italy had tried the same, and been startled by the result. So he wrote me, and I thought, “That can’t be right”, but the experiment replicated. Crazy.
Now, it’s tempting to jump straight into litigating the blog post itself. Maybe you like wolves, and think 1,285 dead sheep, and exponentially counting, is an acceptable price to pay for their presence. Maybe your suicidal empathy weighs the plight of Roma migrants above that of Copenhagen residents who say they’ve had enough. That’s fine. Reasonable people can disagree on what it takes to maintain a functioning society.
But the idea that Anthropic should decide what you get to read is insane. Translating a text isn’t endorsing it. What if Microsoft Word started refusing to right-align paragraphs it disagreed with? Or Windows wouldn’t print an excerpt from The Camp of the Saints? (I know, don’t give them any ideas!).
