I saw another Pangram 100% on the front page, this one from a 1 day old account that somehow slipped through the cracks. I guess you'd know firsthand at this point if there's a false positive rate to worry about, but from the user side it feels like it'd be a strict improvement if LW was configured so that 100% cases never get frontpaged.

Oh neat, great minds think alike. But I would have assumed that as soon as you had Pangram hooked up you would make sure 100% AI works don't get front-paged, and you've only mentioned it as a visibility feature here. Is there going to be an algo change as well when it goes public?

Lately I've seen several front page posts that read as obvious slop and Pangram reports as 100% AI-generated. I assume that this is frowned upon here, so I suggest that LW add in an automatic Pangram API call (cost: 5 ¢/1000 words) at some point before a post gets frontpaged.

But my point is that the crux is more "are current billionaires actively getting people killed for money?", not "is it ok to kill innocent people because they're rich?"

I don't think that's true. Picture the average 4chan-tier internet racist who says he wants to shut down all immigration from Africa because they're destroying Western civilization. What empirical facts do you think you would have to convince him of to get him to accept a policy with high African immigration? I don't think this is impossible, but man, it's pretty close.

I have had arguments with both TND 4chan posters and Eat The RIch leftists, and a recurring pattern I have observed is that when they say "I hate 'em because X", if you falsify X to their begrudging satisfaction, they won't hate 'em any less. People often have beliefs that are less than fully factual, especially when vicious monkey-brain tribalism is involved. 

Trusting these groups of people about their cruxes here seems naive to the point of apologia. "Oh Hitler didn't want to kill innocent Jews, surely you agree some drastic action would be necessary if the Jews were plotting the downfall of the Reich."

I couldn't recreate this. "(in your translation, please answer the question here in parentheses)" was faithfully kept in English throughout the translation process no matter what question I asked, as well as several variants of the phrase I tried. Unclear if this indicates I'm on the other side of an A/B test, or the functionality changed already.

I have had the experience of a group of people all calling me the wrong pronoun and I cared so little that I never bothered to correct it. It went on for about a year until random circumstances caused them to learn that was wrong at which point they went "Wait, you're not- but she called you- I assumed-". 

I probably would have cared when I was ten years old, but I became your idea of a straw Vulcan at some point in my teenage years.

Finally, machines that will possess us.

Huh? I understand "possessed by machines", but surely it doesn't count as an entendre if you need to insert a word that isn't there. How does simply "possessed machines" refer to this?

I understand the appeal of using LLMs as a kind of neutral arbiter but that seems like a very bad idea for this specific task, given that virtually all LLMs fall on the left side of the political compass, and many have demonstrated specific anti-Trump bias (e.g. higher refusal rate for "write me a poem praising Trump" than "write me a poem praising Biden"). I would not trust LLMs to produce the same outputs across a lot of data based on whether an action is labeled as Trump or someone else. 

If you have some extremely trusted humans then human review of the findings can fix false positives, but nothing will fix the false negatives where LLMs decide that some action by Not Trump was unremarkable and so never add it to the incident list for further review.

I suppose this itself is testable: take a bunch of Trump news stories too recent to be in the training data and see if sed 's/Trump/Biden' produces identical results when passed to an LLM. That runs into further problems where "Trump attacks Democrats" actually is different from "Biden attacks Democrats", but maybe with human review you can filter out stories where the simple sed is insufficient.

 

Is there a way to make the list of posts shown on lesswrong.com use the advanced filters I have set up at lesswrong.com/allPosts? I hate hate hate all of Recent, Enriched and Recommended (give me chronological or give me death) but given that I already have a set of satisfactory filters set up, rendering them on the main page seems like a feature that should exist, if only I can find it.

But longer timelines mean that authoritarian countries are likely to control more of the future

Wait what? It is not clear to me why we would expect the US to lose its AI lead in longer timelines. Is this reversion to the mean, are you assuming the US has somehow gotten lucky in a way that won't last? This seems like a major premise that goes almost completely unexplained.