Ive come to appreciate pride as my favorite month of the year as i can scroll twitter for literally hours just looking at these "women" and laughing hard at them. I've probably bookmarked over 200 of these today alone.
One of the weirder things about LLMs is that they can know a fact in one context, but not in another. How does that work across languages? If you give models the same benchmark tests in French or Latin, does it score as high?
I'm not surprised. And I'm *assuming* some of that leaks from English, or rather than a lot of AI's rhetoric happens before it gets concretized into a language, and English is the source of that rhetoric. But if so, then we should have LLM-isms that come from other language languages, like Chinese.
People who know: Do LLMs have different "typical AI writing" styles in different languages? What are they? Can someone look at some Latin and say "Ugh, that's ChatGPT. Too much 'non solumโฆ sed etiam,' or whatever"
An internal poll showing +4 is not super reassuring given that internal polls typically exaggerate their candidate's standing by 4 points or so. And that's smaller than Platner's lead in most public polls before.
This worked in Star Wars. You got a sense that the world was evil but normie evil, and Vader didn't fit in. He didn't wear a uniform! He boarded ships and piloted a fighter! He was an evil relic. As Tarkin put it "The Jedi are extinct. Their fire has gone out of the Universe."
The later movies undid all of this. They made the Emperor a force-wielding wizard. The Emperor should have been like Tarkin.
This worked in Star Wars. You got a sense that the world was evil but normie evil, and Vader didn't fit in. He didn't wear a uniform! He boarded ships and piloted a fighter! He was an evil relic. As Tarkin put it "The Jedi are extinct. Their fire has gone out of the Universe."
The later movies undid all of this. They made the Emperor a force-wielding wizard. The Emperor should have been like Tarkin.
No, I completely agree. Andโembiggening your tweetโthat's my feeling about LLMs generally: Some good uses but a *disaster* for education. I could see a future where they do much good and enhance productivity considerablyโand they're still a net loss, because they fuck up kids and education. And all this on top of the damage done by smartphones!
Its rewrite:
LLMs are basically the median reader. Useless for academic writing or literary fiction; perfect for blog posts and marketing copy, where the median reader is the whole audience. Great at flagging when I've switched terms or gotten too clever with a word โ and great at killing my darlings.
IT GAVE ME AN EM-DASH! ๐
There's a difference between using AI and using AI to write your article. One could, for example, hardly write an article about AI systems without *using* AI! But one should not not use AI to write the ensuing article; doing so is disrespectful of your readers and only illustrates your intellectual defects.
As for quantitative fields, humanities is rarely quantitative, and you were talking about humanities. But the important point is about writing.