I thought it might be “look for things that might not even be there as hard as you would if they are there.” Then the koan form takes it closer to “the thereness of something just has little relevance on how hard you look for it.” But it needs to get closer to the “biological” part of your brain, where you’re not faking it with all your mental and bodily systems, like when your blood pressure rises from “truly believing” a lion is around the corner but wouldn’t if you “fake believe” it.
danielechlin
Karma: 59
Neat. You can try to ask it for confidence interval and it’ll probably correlate against the hallucinations. Another idea is run it against the top 1000 articles and see how accurate they are. I can’t really guess back-of-envelope for if it’s cost effective to run this over all of wiki per-article.
Also I kind of just want this on reddit and stuff. I’m more concerned about casually ingested fake news than errors in high quality articles when it comes to propaganda/disinfo.
By “aren’t catching” do you mean “can’t” or do you mean “wikipedia company/editors haven’t deployed an LLM to crawl wikipedia, read sources and edit the article for errors”?
The 161 is paywall so I can’t really test. My guess is Claude wouldn’t find the math error off a “proofread this, here’s its sources copy/pasted” type prompt but you can try.
You want to be tending your value system so that being good at your job also makes you happy. It sounds like a cop-out but that’s really it, really important, and really the truth. Being angry you have to do your job the best way possible is not sustainable.
“Wrap that in a semaphore”
“Can you check if that will cause a diamond dependency”
“Can you try deflaking this test? Just add a retry if you need or silence it and we’ll deal with it later”
“I’ll refactor that so it’s harder to call it with a string that contains PII”
To me, those instructions are a little like OP’s “understand an algorithm” and I would need to do all of them without needing any support from a teammate in a predictable amount of time. The first 2 are 10 minute activities for some level of a rough draft, the 3rd I wrote specifically so it has an upper bound in time, and the “refactor” could take a couple hours but it’s still the case that one I recognize it’s possible in principle I can jump in and do it.
Allow grade skipping
I get you’re spitballing here, and I’m going to admit this isn’t the most data-driven argument, but here goes: you’re saying take away the kid’s friends, ratchet up the bullying, make sure they hit puberty at the wrong time, make sure they suck at sports, obliterate the chance of them having a successful romantic interaction, and the reward is one of two things: still being bored in classes with the same problems, or having “dumb kid” problems in the exact same classes that are harmful to dumb kids.
Again… total leaf-node in your essay and sometimes “bad” ideas can just expand imagination for what sort of outside-the-box solutions we should consider. Felt this was worth calling out as a uniquely bad idea.
Also I did this, I skipped 8th grade. Which was silly. I was actually hit-or-miss behind in humanities classes, for a similar reason you call out in math: no teacher noticed I had a mental process error where I didn’t know the difference between reading and processing a sentence and unpacking the sentence’s meaning, and The Odyssey was hard enough to show that difference. (Today, honestly, I still can’t follow the plot of Elden Ring and Christopher Nolan movies and stuff, but I’m very good at themes and theses once I know what’s going on.) I would have needed, yeah, something around 4 years jump to be stimulated in a math class again, which is actually what happened at PROMYS my sophomore-junior summer. PROMYS was somewhat socially smoothed out too. I was still young-ish but more on the bell curve in age instead of just youngest in my class by a full year.
So I pretty much just think of high school as a black hole in my life plot, between a more-or-less happy childhood and more-or-less happy college experience, with some sun peaking out (mixed metaphor sorry) during PROMYS. And that’s how I described it in high school too. This should sound sad, and it is sad, because just finding some creative way to turn 4 years of youth into something no better than sitting in a waiting room a little too long (and that’s already the upgraded “alternative thought” in place of thoughts of self-harm or something) is a bad ends, even if I’m a happy-ish worker bee now.
If I had to offer an alternative proposal. It seems that remote learning is, you know, bad socially, but maybe a little strategic remote learning here and there to accelerate someone in math or get them reading harder fiction sooner, across a larger pool of kids, isn’t all bad? Like are we going to humor the thesis that being bored in the same classes is key to making friends at lunch? If you think so, the problem is probably one kid being an isolate, not the system just working that way normally.
You also have the Trump era “RINO” slur and some similar left-on-liberal fighting. First let me deal with this intra-tribe outgroup, before getting back to my normal outgroup.
Whenever I try to “learn what’s going on with AI alignment” I wind up on some article about whether dogs know enough words to have thoughts or something. I don’t really want to kill off the theoretical term (it can peek into the future a little later and function more independent of technology, basically) but it seems like kind of a poor way to answer stuff like: what’s going on now, or if all the AI companies allowed me to write their 6 month goals, what would I put on it.
Camus specifically criticized that Kierkegaard leap of faith in Myth of Sisyphus. Would be curious if you’ve read it and if it makes more sense to you than me lol. Camus basically thinks you don’t need to make any ultimate philosophical leap of faith. I’m more motivated by the weaker but still useful half of his argument which is just that nihilism doesn’t imply unhappiness, depression, or say you shouldn’t try to make sense of things. Those are all as wrong as leaping into God faith.
It’s cool to put this to paper. I tried writing down my most fundamental principles and noticed I thought they were tautological and also realized many people disagree with them. Like “If you believe something is right you believe others are wrong.” Many, many people have a belief “everyone’s entitled to their own opinion” that overrules this one.
Or “if something is wrong you shouldn’t do it.” Sounds… tautological. But again, many people don’t think that’s really true when it comes to abstractly reasoned “effective altruism” type stuff. It’s just an ocean of incomprehensible rightness and wrongness, better to focus on yourself and your loved ones.
“Corruption is bad.” Nick Rafter recently had an article showing how your prior should really just be that elites manage corruption and normal people really don’t care, they just want results.
“Belief is both rational and emotional/psychological/biological”… yep, that angers people on both sides.
This might just be a writing critique but 1) I just skipped all the bricks stuff, 2) I found the conclusion was “shares aren’t like bricks.” Also like what should we use instead?
I’ve been making increasingly more genuine arguments about this regarding horoscopes. They’re not “scientific,” but neither are any of my hobbies, and they’re only harmful when taken to the extreme but that’s also true for all my hobbies, and they seem to have a bunch of low-grade benefits like “making you curious about your personality.” So then I felt astrology done scientifically (where you make predictions but hedge them and are really humble about failure) is way better than science done shoddily (where you yell at people for not wearing a mask to your intensity.) So I settled on the 52⁄48 rule—science, the truth, liberal democracy, all of these things have about a 2% edge over their enemies. It’s very rational to wind up in the 48 (a small mistake not a big one) and very hard to beat a 48 when you need to (like persuading people to take vaccines). I agree that humility is a good start. This seems to fit what I’ve lived through much better than my old ideology prior of like, 100-epsilon/epsilon.
I know I’m writing in 2025 but this is the first Codex piece I didn’t like. People don’t know about or like AI experts so they ignore them like all us rationalists ignore astrology experts. There’s no fallacy. There’s a crisis in expert trust, let’s not try to conflate that with people’s inability to distinguish between 1% and 5% chances.
Reminds me of the Tolkien cosmology including the inexplicable Tom Bombadil. Human intuition on your conjecture is varied. I vote it’s false—seems like if the universe has enough chances to do something coincidental it’ll get lucky eventually. I feel that force is stronger than the ability to find an even better contextualized explanation.
I almost think it’s a problem you included the word “mainstream.” It’s a slippery word that winds up meaning “other people’s news.” It seems like realizing the point in your post is one step, and taking a more surgical dive into what news counts as obscure enough is another. If you’re a doomscroller you’re probably gravitating toward stuff many people have been hearing about, though.
The “semantic bounty” fallacy occurs when you argue semantics, and you think that if you win an argument that X counts as Y, your interlocutor automatically gives up all the properties of Y as a bounty.
What actually happens is: your interlocutor may yield that X technically counts as Y, but since it’s a borderline example of Y, most of Y doesn’t apply to it. Unfortunately, as the argument gets longer, you may feel you deserve a bigger bounty if you win, when really your interlocutor is revealing to your their P(X is not Y) is quite high, and if they do yield, it’s more likely they’re yielding that X is a borderline Y.
This is a “leaky generalizations” and more specifically a “noncentral” fallacy. It happens, for instance, when someone tries to prove something is racist so as to imply behavior should change and gets resistance. In fact, you might consider that gap between “technically a Y” and “typically a Y” as a sort of semantic deficit. If the point of something being in Y is that it’s bad, consider arguing that it’s bad without even bringing up Y.
This applies even to some of the most drastic, moralized words we have, like “slavery,” “genocide” and “fascism.” However you feel about any issue in one of those topics, I will inform you that proving to someone that that thing is in that category, is not going to have the effect you want. There is no semantic bounty.
Well a simple, useful, accurate, non-learning-oriented model, except to the extent that it’s a known temporary state, is to turn all the red boxes into one more node in your mental map and average out accordingly. If they’re an expert it’s like “well what I’ve THOUGHT to this point is 0.3, but someone very important said 0.6, so it’s probably closer to 0.6, but it’s also possible we’re talking about different situations without realizing it.”