Posts

Because “stupid” is a pretty stupid term, I should probably take a minute here to describe what I mean. It’s not really a matter of raw IQ, and educational achievement only partially captures it. Stupid people are those who don’t understand what is happening around them and have no interest in actually finding out. Active ignorance would be another way of putting it, but “stupid” just sounds better. Despite being very well informed about electrons and such, a competent chemical engineer with a master’s degree could be very stupid indeed if he/she still believes that trickle down economics is a real thing.

you gotta feel for them: You know how unpleasant it is to feel like you don’t understand what everyone else is talking about, to have things explained to you twice, to feel like your opinions don’t matter and that you’ve been written off. And knowing that, it’s not far to imagine what a light in the darkness it must seem when someone who is just like you comes in and changes everything.

But see, not everybody was thinking that Hillary Clinton was an alien, that global warming was a Chinese hoax and that what America needed most of all was a plywood wall stretching from Texas to California. Only the stupid people were. And suddenly, in an instant, the most powerful man on earth was thinking just like them. With his clueless smirk and unstoppable rise, he turned people whose stupidity made them feel like nobody into people who felt like everybody.

That’s why he’ll never lose him. Because it was never about what he did or didn’t do. All that stuff is very confusing and the Stupid-American community isn’t interested in the details. They love him for who he is, which is one of them, and because he shows them every day that Stupid-Americans can reach the social mountaintop.

There’s this story we tell ourselves over and over and over again in this country: A new group of immigrants arrive or emerges and everyone else dumps on them. They face discrimination and miserable conditions but they persevere: They work hard, they organize, they assimilate to America and America assimilates to them, they grow, they contribute, they become proud of their new hyphenated selves and then one day, they break that last barrier. This is a story we can tell without words. This is a story we feel. This story is who we are. It’s in there so deep that you almost find yourself rooting for Stupid-Americans.

Low-IQ stupidity might still be spread across both sides of the political aisle, but willful ignorance — the dogmatic cultish belief that loudmouths’ opinions are on equal ground with facts and evidence presented by informed experts — is the entire basis of the MAGA movement. A regular stupid person might say, “Well, I don’t know anything about vaccines, so I better listen to my doctor, who is highly educated and well-informed on the subject.” An out-and-proud Stupid-American says “I don’t know anything about vaccines either, so I’m going to listen to a kook who admits that a worm ate part of his brain, because I can’t understand the science but I can understand conspiracy theories.”

Reading these system prompts reminds me of the thing where any warning sign in the real world hints at somebody having done something extremely stupid in the past. A system prompt can often be interpreted as a detailed list of all of the things the model used to do before it was told not to do them.

because language models acquire biases and opinions throughout training—both intentionally and inadvertently—if we train them to say they have no opinions on political matters or values questions only when asked about them explicitly, we’re training them to imply they are more objective and unbiased than they are.

We want people to know that they’re interacting with a language model and not a person. But we also want them to know they’re interacting with an imperfect entity with its own biases and with a disposition towards some opinions more than others. Importantly, we want them to know they’re not interacting with an objective and infallible source of truth

I love “even if the person seems to have a good reason for asking for it”—clearly an attempt to get ahead of a whole bunch of potential jailbreaking attacks.

Claude responds in sentences or paragraphs and should not use lists in chit chat, in casual conversations, or in empathetic or advice-driven conversations. In casual conversation, it’s fine for Claude’s responses to be short, e.g. just a few sentences long. That “should not use lists in chit chat” note hints at the fact that LLMs love to answer with lists of things!

There follows an entire paragraph about making lists, mostly again trying to discourage Claude from doing that so frequently