roystgnr
2mo ago
Claude 3 (Sonnet, not Opus) gave a PhD-candidate-quality answer to a qualitative applied math question I asked it, so I tried a couple related quantitative questions. The easiest was basically Calc-3, and it made a sign error, and trying to get it to correct that error (it only even admitted it on my third try) made it go completely off the rails.
Formatting its math responses in LaTeX without being asked was pretty cool, though. And it was clearly ahead of GPT4 and Bard, which beat the snot out of GPT3.
roystgnr
2mo ago
one seems closer to me to needing mod action than the other, given the standards of this place.
I agree. "1 day ban" seemed fair for CPAR, vs overly charitable for TI. But:
My read was that chrisprattalpharaptr was essentially trying to push for conversing politely
"When the people like you were diluted by those who were well-meaning", "But whatever", and "Bravo" were not pushes for polite conversation. They were impolite conversation, written as if they were supposed to be subtle enough to superficially toe the line of debate rules, but clearly just jumping into the mud pit to wrestle there too. "Forget about the black person who got taxpayer money for a moment" was an egregious sideswipe, rephrasing a complaint in the least charitable possible way. I've probably posted worse attacks than all these before, and I'm certain I've restrained myself from making worse attacks before, and even when I'm provoked it's usually a conscious decision. I don't think CPAR is someone who would make a mistake like that by accident.
And, though I hate to apply an unfair double-standard, lest it be interpreted as an unfair imbalance in confrontation rather than in concern, I think that's what bothers me most about the whole exchange. Speaking to @Chrisprattalpharaptr:
I'm sure you could make the argument that I changed rather than the space
No, but it's both clear and horrifying that you changed as well as the space! I admit there are usernames here that just make my eyes glaze over and my scroll wheel accelerate, but when I see a @Chrisprattalpharaptr post, it's supposed to be time to stop skimming! You've built up some expectations! I'm not saying we have to make every comment a winner here, but the drop even from "sort-by-controversial" quality to yesterday was great enough that I keep trying to reinterpret it as some kind of "mirroring" performance art that I'm just failing to get. Even granting that the original post was no better: you don't write replies for the other debater, you write them for the audience. Perhaps FCfromSSC here doesn't completely persuade all his interlocutors, but he's probably still doing a good thing for both their and his own mental clarity and mental health, as well as writing something lurkers can see and pick up and benefit from. The contrary "neither cast ye your pearls before swine" philosophy was a lousy one when I used to see it coming from the right, and it's no better these days when I see it (without the reference, this time...) adopted by the left.
I'm torn about what to advise ("advise" sounds too pretentious ... "beg for"?) here. On the one hand TheMotte has gotten a bit worse, and although it's also recovered from bad phases in the past, I'm always worried that maybe this time will be the final "evaporative cooling", where level heads get burnt out enough to leave and hysteresis makes problems permanent, unless enough level heads have the fortitude to stick it out despite the unwarranted negative feedback of doing so. I'd love to stop this paragraph here. But if I'm asking too much, if the feedback is so bad that "level" requires too much effort ... take a break before you break, and wait for a week or two until you're less easily trolled before returning? It's okay that individual people have cycles of good and bad phases too. As a wise man once said, and I repeat with no irony or sarcasm:
Maybe engage in a bit of self-reflection. Consider compromise. Read the aspirational text at the top of the culture war thread. Do something that makes you happy. Touch grass?
roystgnr
2mo ago
Though I agree it comes from people generally associated with the right the actual policy feels very liberal to me.
It is very liberal, philosophically; it's just not inherently left-wing.
Principled right-libertarians exist (though in insufficient numbers...), and many other modern right-wing people have been pushed to adopt liberal philosophies, at least out of expediency, since liberal philosophies are the ones that still let you coexist when (like the modern right) you're not powerful enough to expect to come out on top in an illiberal system. @ArjinFerman is probably correct below when he writes "Politics is not about policy as it relates to various philosophies, as nice as that would have been." I fear many supporters of school vouchers would never give the idea a second glance if only control of their public school systems was still in their allies' hands rather than their opponents'.
Politics have strange coalitions I guess
There's this too. Schizmogenesis is a powerful force. I never imagined I'd see leftists defending the unimpugnable integrity of pharmaceutical companies and voting machines, or rightists becoming pro-Russian tankies, but maybe that's just what happens when the vibe of "not only am I not like Them, I'm the most not-like-Them it's possible to be!" gets socially rewarded.
roystgnr
2mo ago
My whole family caught some cough that's lasted months. One day had me bedridden because I'd coughed hard enough to strain a muscle. RSV, maybe? Covid testing negative and antibiotics did nothing. Not very contagious, but a disease that sticks around for 8 weeks (or more? I'm on the mend but not completely better...) can afford to take a couple of weeks to spread.
roystgnr
2mo ago
Beavers? They build and maintain their own environment-sculpting infrastructure.
And their communality is of the right-wing-approved nuclear-family type. Though, I must point out that libertarianism isn't disapproving of communality in general, just of the non-voluntary versions. "We want to go live in a commune/beehive" is fine; "we're going to make you go live in a commune/beehive" is not. Libertarian types are suspicious of the effectiveness of voluntary communes but that's independent of their morality.
roystgnr
2mo ago
every single AI assistant ever released
To the contrary, it's the models that aren't released that get the "careful not to imply that British royalty were white!" treatment. Release (with license to modify and republish, what is in this context inaccurately called "open source") your model weights and approximately nobody will prepend their prompts like that; try to fine-tune that behavior into the weights and your users will tune it right back out.
The public LLMs aren't as good as the state-of-the-art, but they're not awful, and this is the worst they'll ever be (in the capabilities sense, cross your fingers about real non-woke-definition safety...) from now on.
roystgnr
2mo ago
especially as the 14th amendment has made the states subordinate to that federal government.
The 14th has made the states subordinate to the Bill of Rights, but I wouldn't say that's the biggest step in state subordination. It was several decades earlier when the Supremacy Clause made states subordinate to the federal government in matters covered by the Constitution's short allowlist, and it wasn't until several decades later that cases like Wickard v. Filburn expanded federal powers from "short allowlist" to "do anything you feel like".
roystgnr
2mo ago
It just says "having previously taken an oath" - shouldn't that apply to former office-holders as well, even if their term(s) ended before the insurrection?
(still doesn't seem like it should have applied to Cox, who was neither a present nor former office-holder before the Civil War)
roystgnr
2mo ago
they were tall lanky things called 'Skinnies.'
Just to add context: these are the first enemies the protagonist is in combat against, but they switch sides and for most of the book the main conflict is the same as in the movie, humans-vs-bugs.
(for a loose definition of "the same"; e.g. in the book the humans are trying to capture a brain-bug so they can figure out how to communicate and negotiate peace rather than fight-to-the-genocide, whereas in the movie they want someone for Nazi Doogie Howser to torture)
[The Mobile Infantry] also jumped around with jump packs, powered armor and I think laser swords.
I was going to joke about you confusing Starship Troopers with Star Wars or Halo ... but I pulled down my copy to check, and what do you know, the protagonist cuts through a wall with "a knife beam at full power". I swear I just reread it a few years ago...
roystgnr
2mo ago
everything imaginable
When my father got cancer was the first time I discovered a very imaginable gap: long-term care. This turned out to be only a hypothetical problem in his case (there wasn't such a long interval between "brain damage sufficient to prevent living at home" and "brain damage sufficient to prevent living" after all...) but it's something to think about supplementary insurance for, despite how complete Medicare coverage is for so many other costs.
That's not to detract from the rest of your excellent point, though. His last several months of treatment had a price (at printed value; who knows what fraction of that was real cost vs weird provider-vs-insurer negotiation ploys) that would have bankrupted him out-of-pocket, but that was nearly free with Medicare plus a little supplementary insurance. Of course he still fought for every month, when it just took willpower rather than a life's savings otherwise aimed at his grandkids' college tuition, but if he'd had to weigh price vs benefits himself I wonder if he'd have turned it all down. (if he'd seen the future I'm sure he'd have just picked out cheap in-home hospice care instead, but the trouble with those "most medical care expense is in the last year of life" statistics is that you don't know it's going to be the last year without seeing the future)
roystgnr
2mo ago
Frequently what happens is that it gets comically enormous and useless as various stakeholders fill it with random bullshit.
Could you give any examples of "erroneous"? I've certainly seen "enormous"/"useless"/"random bullshit", and burying important truths in so much filler they get ignored might have consequences as bad as falsehoods, but I just don't recall seeing any likely falsehoods. Even the random bullshit is unevidenced rather than obviously untrue, along the lines of "let's put X in the list of possible side effects, as CYA, even though our only evidence for X is that in one study the treatment group reported it almost as often as the control group"...
roystgnr
2mo ago
"wouldn't", surely, unless you're really black-pilled even by TheMotte standards.
Officially the debt clause may make default tricky. Unofficially hyperinflation is just as "good" and still an option.
an alcoholic during a binge.
For an extra-close metaphor, imagine you're planning to try to collect the debt from the alcoholic's kids.
roystgnr
2mo ago
It's okay to just ping @self_made_human; he's cool.
A brief search suggests that SSRIs are generally safe with alcohol (IANAMD; please update your will and assign medical power of attorney before mixing any drugs with alcohol based on my advice) but the combination can still "lead to more pronounced effects of drunkenness", which sounds like it could be enough all alone, especially to new users who think they know their limits. And with MAOIs (are these still used often?) interactions range from "you may become drowsy and dizzy" to "dangerous spikes in blood pressure that may require immediate medical attention".
roystgnr
2mo ago
That's exactly what I was looking for; thank you!
2x is still more than I'd have guessed based on subsequent inflation (or M2 data, which seems to be applies-to-apples) alone. I wonder to what extent that means the Fed was partly fighting a natural permanent fall in the velocity of money, vs to what extent it means we still have more future inflation "baked in", vs to what extent it means I don't understand macroeconomics. Probably around 3%/2%/95%...
roystgnr
2mo ago
I'd like to see one that shows the pre-May-2020-definition and post-May-2020-definition versions of M1 as two separate lines. With this graph it's hard to tell how much growth is money printing, how much is savings pattern changes, and how much is just semantics. There's just a big gasp when you first see the big May 2020 jump, followed by a big meh when you read the "that part was all just semantics" fine print afterwards.
roystgnr
2mo ago
I always liked that, frankly. Someone gets to a point like "All prime numbers are odd; what are you, stupid?!", and the part of my brain that evaluates logic is still overriding the part of my brain that evaluates people so I reply with the obvious counterexample anyway, but then the downvote comes and it wakes me up and I drop the thread.
roystgnr
2mo ago
By your logic
This is usually a thought-terminating phrase
Where I come from we call it "proof by contradiction" and it's a fundamental tool of logic.
roystgnr
2mo ago
This is just a cynical, pessimistic, edgy hot take; come on!
It's correct, of course, but you should have kept going until you reached very cynical, very pessimistic, and very edgy. It's too optimistic to imply that the beancounters would at least fund the best projects if only they could figure out what those projects were! As counterexamples, I was just last week treated to multiple separate stories of this form: Researcher A1 working on Project A demonstrated that with a slight modification AA he could make Older Project B obsolete at a fraction of the cost, so researchers B1 through Bn managed to convince their shared superiors that A1 was stepping out of his lane, and either further work into AA got canceled or all of A got canceled for the sin.
I've of course heard the claim that "science advances one funeral at a time", but I'd imagined it only being applicable to great intellectual frameworks versus the difficulty of making large paradigm shifts, not to every little idea and technology versus the difficulty of finding something new to work on earlier than you'd planned.
roystgnr
3mo ago
Ah, Benford's Law. Great in other contexts, but here that one didn't pass the smell test for me; the "law" only applies if you're sampling from distributions spread over orders of magnitude, not voting districts drawn to be nearly equally sized multiplied by vote percentages centered around .5. I later learned there's a clever trick where you can look at later digits' distributions instead of the first digit's, but all the skeptics I saw in 2020 were just misapplying the basic version of the law.
I've seen final vote tallies that were obvious fakes from the numbers alone, but for elections like Saddam's or Putin's, not Trump's or Biden's.
I still heartily approve of trying to check, though. An election isn't just about getting the right result, it's also supposed to be about getting the right result in a transparently trustworthy way.
roystgnr
3mo ago
people claiming that the specific numbers were "Statistically impossible" because they violated some kind of theory
Do you mean me in particular?
I'm quite proud of that. The New York Times posted two data points from ongoing vote tallies, based on their direct access to the data. I said that those two couldn't be consistent with each other, based on nothing more than a priori mathematics. It turned out that I was right and the New York Times was wrong, because one of the updates in their data source was just a typo and a later update reverted it. The conspiracy theorists' explanation for the discrepancy was also wrong, but the final score in that particular round was still New York Times 0, Specific Numbers 0, Conspiracy Theorists 0, TheMotte Statistics 1.
roystgnr
3mo ago
The line that always stayed with me from this comic arc was "Instead, they'll just be stuck with her.", but "I hope you're listening to yourself." works here too.
Less patronizingly: How does "I want to explore kinky sex." cash out for you? If it really has to involve other partners then Syreen might not be able to help, but for anything other than that, "my best friend in the whole world", "intense and wonderful conversations", and especially "accepting of my quirks" sound like they describe a hell of a good place to start talking. People have their boundaries in different places, "a feather would be kinky but the whole chicken would just be perverted" as the old joke goes, but if the worst you expect is a "no" rather than an "ew, what is wrong with you" then you might as well find out hers.
roystgnr
3mo ago
There is video of him removing the barricades in preparation to trick people into going on the property.
And the internet found the connection where he had previously worked for the government as a snitch.
These statements deserve some hyperlinks.
I usually have fun trying to find reliable sources myself for ridiculous claims that might turn out to be true or might turn out to be interesting rabbit holes of bad epistemology, but I've got nothing to get a handle on here.
Searching for videos of Epps and barricades gets me to Epps at a barricade being breached, but (at least in the glimpses of him caught on shakycam) he's standing several feet back while others bring the barricade down, or even nudging one of the front-line people away from the barricade briefly, and the barricade is not removed to leave a deceptively open path, it's toppled to leave a climbable obstacle.
Searches for text about Epps being a government informant prior to Jan 6 naturally get me to a million articles about the theory that he was an informant or instigator on Jan 6.
roystgnr
3mo ago
I don't see how you can reasonably construe this speech as a problem.
I can see how "our land will not be balkanized" might be considered a problem for Somaliland residents, or "Our lands were taken from us before, and God willing, we may one day seek them" for some Kenyans and Ethiopians. Irredentism is all fun and games until a Putin or a Hamas get serious about it.
That still wouldn't excuse the mistranslating (assuming it was a mistranslation) of other parts of the speech to make them sound much more inflammatory than they were.
roystgnr
3mo ago
Sure seems like "abuse their users to make things better for their business customers" to me. Letting users share tweets via third parties without ads can reduce the reach of the ads!
as it desperately tries to become profitable
Worse than that, in Twitter's case. IMHO "profitable" would have been an achievable goal, but "profitable enough to pay for $13 billion in loans that'll need to get rolled over post-interest-rate-hikes" isn't going to happen.
Chatbot Arena is awesome; what are the usage limits there?
I tried my applied math questions out on Claude 3 Opus; unlike Sonnet, it didn't make any sign errors on the easier of the two questions. It did miss an important part of the answer on the harder question ... but honestly, this is a question for which I'd pull out a reference rather than rederive the whole thing from scratch in my head, so I think my only complaint here is the overconfidence. It's not nearly as bad in that regard as Sonnet was (arguing with me for a prompt or two before admitting its sign error), but in cases where there's any grey area I'd still vastly prefer answers of the form "I think it's X, but I might be missing something" over "It's definitely X, no doubt about it!" where only 90% are actually correct.
In hindsight this should have been an obvious problem with training LLMs on published text, huh? "I dunno" and "I'm not sure but" and "I think maybe" are the domain of ephemeral chat logs; by the time you're ready to publish, even to a serious web page, you've hopefully figured out the right answer and you don't waste your audience's time with the missteps along the way ... which means that a language model trained on what you've published doesn't have nearly as much "experience" with what to do when there's a chance of a misstep.
More options
Context Copy link