roystgnr
11mo ago
We could get schools to turn out 0% failures, by mandating that no child ever be given an F. This would work in the same sense, and backfire in the same more important sense, as encouraging children to be smart and engaged and curious by voting.
That said, a supermajority of adult voters are barely doing anything I'd recognize as voting, so I'm all in favor of letting the kids have a shot too. There might at least be a few years' period during which they vote based on the "study the candidates and pick the best one" ideals, before "You can't vote against our team's corrupt handsy geriatric, or their team's corrupt handsy geriatric might get in!" messaging catches up to them.
roystgnr
11mo ago
It would make sense for everybody to have those feelings since Omicron, when we were suddenly topping 5 million confirmed new cases per week in the US. I was quite paranoid about avoiding Covid before the vaccines came out, and even afterward I figured it couldn't hurt to stay masked in public and reduce transmission rates a little and get exposed less frequently ... but after the peak outbreak case rate jumped 5-fold? At that point it was clear we're all going to be getting exposed to Covid again, frequently, for the rest of our lives, unavoidably, and "just hope your immune system can handle it each time", maybe with a slight improvement via "don't get overweight", is the last defense.
roystgnr
11mo ago
My kid starts Algebra 1 next month, and she's 9. That's not prodigious, but it's pretty good.
It's not Von Neumann or Tao, but it's definitely child prodigy level. Recall that the state of California has treated "we can't really teach Algebra I to the top 13 year old math students" as if it's a serious proposition to debate and not just grossly unfit educators outing themselves.
roystgnr
11mo ago
I don't think you should be able to show a firearm unless you're in a place / circumstance where you can legally discharge that firearm.
Could you clarify? The place you've described is "any place", which doesn't disallow non-uniformed shooters, and the circumstance you've described is "with justifiable belief in an imminent unprovoked threat of death or grievous bodily harm", which does disallow police officers and security guards at any time before it's too late to go get a gun.
And it makes it so that if you see someone strapped, you can know it's a problem and run/call the cops/etc.
Sadly, this rule would only be reliable if certain false prerequisites like "concealed carriers' clothes never shift the wrong way" were true. There are a lot of people who never want to open carry but who also never want to go to jail (or worse; IIRC I read about this in the discussion of a CCW holder killed by police) for not concealing well enough.
roystgnr
11mo ago
a promise that they would receive it again eventually
Is this an interpretation of Brigham Young's "they never can hold the Priesthood or share in it until all the other descendants of Adam have received the promises and enjoyed the blessings of the Priesthood and the keys thereof"? That qualifies as "eventually", and straight from the horse's mouth, but it was at least a bit of a pivot to interpret it not as "after the Resurrection", but rather as "after mid-78".
The church won't pivot on gay marriage.
Maybe not, but if they did it wouldn't be any harder to rationalize. Pivoting on prior scripture about what's clean vs unclean is as old as Christians eating bacon; it even makes sense to the non-religious! Something like "this was actually risky before modern STD cures" (i.e. accounting for some failures of celibacy and trying to minimize the damage) would be closer to "this was actually risky before modern animal husbandry" than to "I guess the pre-existence ran out of less-valiant souls?"
roystgnr
11mo ago
If you shoot at the cops, they know your address.
So "shoot while they're knocking at someone else's door" is the equilibrium strategy, then? At least "one of us got shot before we killed the shooter" can be spun as a heroic story; "one of us got shot before some kid at a desk bombed innocent people" (not to mention the crime scene where evidence of the bullet trajectory used to be) is the sort of thing that makes you look for a better job than "sucker who draws fire on the civilian-bombers' behalf".
And that's assuming no other collateral damage, which is ... a stretch. The 1985 MOVE bombing was horrifying enough to show up in the news last year, even though all the "this is unconstitutional", "pay millions of dollars to the victims", etc. decisions were made decades ago. This does not scale up.
roystgnr
11mo ago
has issued millions of concealed carry permits
That undermines more than supports the argument for permitless carry, doesn't it? I can see a strong argument for permitless carry in states where the legislature says "shall-issue" but the licensing agency says "ooh, sorry, on your application you did/didn't close the top on the digit '4', please try again, that'll be another $200 filing fee", but if training requirements are actually providing training rather than obstruction then they don't seem like a bad idea in theory.
it doesn't cause excess violence or deaths
I don't think you can extrapolate from "Florida allows trained licensees to carry concealed and the homicide rate kept declining" to "Florida allows anyone to do so and the homicide rate won't jump" ... but "dozens of states allow anyone to do so and the homicide rate kept declining" is decent evidence. New Hampshire isn't exactly a murderous hellscape.
roystgnr
11mo ago
GPT-4 can draw (albeit not well) if asked to output SVG or TikZ or some other human-readable graphics format.
I've tried a set of qualitative math/engineering questions on LLMs. The Bard and GPT-3.5 answers were about what I'd expect from an undergraduate starting to study the field: roughly 25% of the answers were true-and-useful, 50% true-but-not-useful (not a bad thing, just statements that were adding context to the answer rather than being the answer), 25% not-true (though even these were oversimplifications and natural misunderstandings, not utter hallucinations). If a human had given those answers I'd have considered them a kid worth mentoring but I wouldn't have expected them to save me any work in the near future.
The GPT-4 answers were better than I'd expect from a typical grad student who had just passed an intro class on the subject. Adequate depth, more breadth, and this time the statements weren't 25/50/25, they were about 75/25/0. I passed my questions to a friend's brother who had a subscription, but now I'm tempted to subscribe myself, give the thing my whole final exam, and see how it does on the quantitative + symbolic questions.
roystgnr
11mo ago
You are pattern-matching to some random commune in Philadelphia.
It's not a perfect match. It's people wanted for illegal weapons possession who got bombed for it, but it was also a group that had been threatening the lives of their neighbors. Many of your future missile victims will be much more sympathetic.
You've found much worse matches, though.
The first key bit with Pearl Harbor was "in port". When Americans died undignified deaths in others' ports, we weren't quite so gung-ho about keeping the pressure on forever. See Vietnam (which had celebrities siding with the Viet Cong, even, not just with civilian collateral damage), Iraq, Afghanistan.
The second key bit was "Japanese". Not just in a racist or myopic "wait, my neighbors' lives matter!" way, but because a coordinated empire trashing our defenses while conquering the Pacific looked like an existential threat. Impromptu snipers would be a threat to the secret police knocking on their neighbors' doors, but nobody's going to imagine that that kid safely behind the drone controls had no other choice.
The last key bit was "mind our own business". Japan had started conquering its neighbors before even economic sanctions started. What is your average hunter doing, that we need to ransack his home if he claims to have lost a gun that you think he's hiding? In this scenario the initial surprise attackers aren't the victims of your missile strikes, they're the perpetrators of them.
Imagine for a moment that we decided to invade Mexico, not because they had knocked out battleships or skyscrapers in a surprise attack, but because they have four times the gun homicide rate that we do and obviously we want to do the most good first, by sending in the military to disarm them all and kill off any resistance. Do you imagine this plan getting wide public support? There may be some "anti-colonialists" who are less resistant to invading Wyoming than Mexico, but I suspect that that group will balk at invading Chicago.
Do you think there will be sympathy for the terroristic gunmen on American media?
For snipers picking off cops that aren't coming for them specifically? No. For victims of misunderstandings ensuing from jumpy cops and drone operators trying to collect guns in such an environment? Absolutely. The largest mass shooting of civilians in US history was committed while trying to round up the victims' guns, and we call it the "Wounded Knee Massacre", not the "Lakota totally had it coming". How many more Breonna Taylor incidents (shot in the crossfire while her boyfriend was shooting at police, yet still the subject of protests for years!) would you expect to see while rounding up the guns owned by ten million African-Americans? How much more extreme would the reaction be if the "crossfire" was a missile and she didn't even have a chance not to die? What about the next time there are kids in the Hellfired house? How about when it turns out that one of her successors had a restraining order against a violent stalker and obviously had a good reason to keep a gun? What about when nobody in the house even had a gun, but it turns out that the cops and military panicked when someone across the street shot one of them from behind? What about when nobody at all had a gun, but some kid set off a firework at the wrong time, or was waving around a toy like Tamir Rice? You're not getting rid of a hundred million guns, even if somehow everybody was on board with that, without triggering a hundred thousand such incidents by accident. And it's going to get worse when terroristic gunmen start triggering such incidents on purpose. There's a quote about the Viet Cong that goes something like: "To demoralize the enemy you send a child carrying a flower and wearing a bomb. To really demoralize the enemy you then send five more children with flowers and no bomb."
"We're going to get rid of guns and shootings by removing part of the Bill of Rights then bringing heavily armed cops from door to door and killing people" is not the obviously easy PR victory you think it is. Have you missed the last few years? At this point "replace as many cops as we can with unarmed social workers because cops can't be trusted with guns" is a serious movement. "Swatting" is a thing you do when you're a horrible person who wants to risk someone's life, not something we want to make mass policy. "Deck the cops out for SWAT and send them door to door" isn't on the table among the left any more, much less the right.
roystgnr
11mo ago
No objections to most of your examples, but:
an increase in violence
(Counter) citation needed? The US reported violent crime rate fell in half over a quarter century, before leveling out. The very recent trend is worrying, but even "let's just ease up on this whole 'police' thing" doesn't seem to have been nearly as disastrous as whatever combination of "let's empty out the asylums" / "let's set a trillion gallons of leaded gas on fire near our kids" / "let's try All The Drugs" / something-else doubled homicide rates in the 60s through 90s.
the only people who seem to exhibit some sort of 'will-to-power' are the radical left, who have insane aims (equity) and insane policies (like depolicing, end of meritocracy, etc)
This is a problem, but is it a major increase? The biggest powers contending for the last century were:
-
FDR's USA: where we shredded the Tenth Amendment to the point where we no longer even realize "United States" is plural, we imprisoned an entire ethnicity, we tried to micromanage the economy with theories as mad as "let's destroy food during the Great Depression", and even our best anti-Depression scheme was the one where we robbed people of gold like cartoon villains. Oh, and we also were lax about nuclear secrets and discharged 85% of our military in the couple years between WWII and the Berlin Blockade, because we trusted "Uncle Joe".
-
Stalin's USSR: the aftermath of radical left gaining power, not just proposing insane policies but killing millions of their own people with them via a sick combination of incompetence and its ensuing scapegoating malice, then doubling down on incompetence by trusting their entire nation to the sanctity of a secret agreement with Adolf freaking Hitler after mutually wiping out the buffer state between them.
-
Hitler's Germany: the very idea of "will-to-power" appropriated from Nietzsche and taken to the extreme, rampaging across subcontinents in an insane attempt to dominate the entire world, but distracted by the obsession with slandering and murdering millions of their own people, and finally defeated in part because ideas like "don't trust that Einstein guy's physics, he's a Jew!" and "let's open up the war on two fronts at once!" are the sort of things you come up with when ruled by your own madness.
I'm not sure how to find political insanity on a graph, and I'll admit it feels like there's been an uptick over the last decade, but we're still nowhere near the heights we scaled during the last century.
I fail to see a universe where AI has total comparative advantage over humans in all things
I do too, but the old "wine and cloth" arguments' conclusions are only ironclad if you ignore both non-labor factors of production and negative productivity.
Non-labor: perhaps "3 guys who bring my coffee+tea+biscuits" have a comparative advantage over 3 hunks of motors and processors doing the same, but if the 3 robots need 2 kWh to do my day's menial tasks whereas the 3 guys need 6000 kcal (7 kWh) per day to stay alive, the robots might still be cheaper. Energy isn't perfectly fungible, there's all sorts of other costs to consider in both cases, etc, but my point is that the fact that humans can always find something productive to do is offset by the caveat that our productivity doesn't just have to exceed zero, it has to exceed a hard floor.
Negative productivity: human productivity doesn't always exceed zero! If reliability (in either positive or normative senses) is required for a job, the negative expected value of potential mistakes and crime can exceed the positive value of the work being done. Even the most haughty CEO-aristocrat has to also be wondering "will humans accept a social structure where the overwhelming majority of them have the status of serfs", for example, right? And if the answer isn't "definitely", then 3 potentially-pissed-off serfs might be too much more of a security risk than 3 potentially-hackable robots. If the upside to human labor is "my biscuits might be seasoned with their delicious salty tears" and the downside is "my tea might be seasoned with rat poison" then the CEO's valuation of human labor might be negative.
roystgnr
11mo ago
The AI doesn't even have to be super-intelligent, it just has to be good at its job.
I think this is one of the creepiest possibilities - that no matter how hard well-aligned independent agentic AGI is, we have to make it soon, because we need something which can think intelligently enough about the A-Z of possible new technologies to say "you'll need defenses against X soon, so here's how we're building Y", independently enough to say "no I'm not going to tell you how Y works yet; that would just let a misanthrope figure out how to build X first", while being trustworthy enough that the result of building Y won't be "haha, that's what kills you all and gets you out of my way" ... and if we don't get all that, then as soon as it's easy enough for a misanthrope to apply narrow "this is how you win at Go" level technologies to "how do we win at designing a superplague" or whatever, we're over.
roystgnr
11mo ago
I'm probably still getting something subtly wrong here. :)
Maybe, but it's at worst an interesting sort of wrong: https://grabbyaliens.com/
roystgnr
11mo ago
Don't forget Nicaragua. Their homicide rate is now down to 7.9/100k: still worse than Texas as a whole, but not as bad as Dallas or Houston in particular, and not even half as bad as such infamous US hellholes as "Philadelphia" or "Columbus".
roystgnr
11mo ago
I still remember marveling that the stupidest-sounding kid in my college engineering classes was getting the highest test grades. It turns out that everybody had stupid questions, and "ask them right away and look stupid" was a better way to handle them than "keep your embarrassing secret and hope you can fix it all by yourself before test time comes".
roystgnr
11mo ago
Super funny written, just confusing spoken. (Unless you're not the one being confusing or confused; then I could definitely imagine this getting padded into a "Who's on first"-quality bit)
roystgnr
11mo ago
Yes, homicides fell.
This isn't the data I linked to. The violent crime rate is about 75x the homicide rate, and both fell in half.
https://journals.sagepub.com/doi/abs/10.1177/108876790200600203?journalCode=hsxa
This talks about a lethality decrease from the 1960s to 90s. I'm talking about reductions in 98% non-lethal crime from the 90s through 2010s.
it's hard to claim the increase in violence as a good thing.
It's especially hard if violence decreased 50%.
roystgnr
11mo ago
Non lethal crime is much more easily decreased by changes in reporting, incentives to report and so on.
This is true. So what happens when we track violence independently of reporting, via victimization surveys? "From 1993 to 2021, the rate of violent victimization declined from 79.8 to 16.5 victimizations per 1,000 persons age 12 or older." The decline is even more astonishing.
roystgnr
11mo ago
The context I always got for "torture doesn't work" was that, while torture works great to get an insurgent to confess that his neighbor is also part of the insurgency, torture is great at eliciting that confession whether it's true or not. If you're lucky you get to parade the neighbor's IED cache out in front of the neighborhood and you have 1 fewer insurgent; if you're unlucky you have to let the suffering neighbor go and you still might have pissed off his further neighbors and cousins and so forth sufficiently to have 5 more insurgents.
Thus everyone had to go to the "ticking time bomb" thought experiment to get a real ethical conundrum: if the tortured suspect is being asked for information where a lie won't hurt any (more) innocents and won't radicalize any more enemies and will be quickly and reliably discovered, then we have to determine whether our values are really enough to say "no".
roystgnr
11mo ago
Homicide isn't tracked by victimization surveys. Unless there's vampire homicides and a particularly brave interviewer, I suppose.
Is the decoupling of homicide from other violent crime during a mass panic something to be really surprised about, though? With 98% of violent crime non-lethal, it only takes a tiny change in conversion rate. If a burglar is suddenly looking at a bunch of Covid-locked-down houses that no longer ever seem to empty, it doesn't seem a priori implausible that a few percent of them are going to say "no, too risky for me" (so the violent crime rate component of robberies still drops) while a few percent are going to say "I need the money, and if it's not empty, I can fix that" (so the homicide rate skyrockets). For that matter, what happens to the other side of the equation during the post-Floyd period? A homeowner who might have said "I'll run and call the police" is now more likely to conclude "the police might just shoot me by accident" or "the police might not even show up tonight" and take things into their own hands. Still a robbery, still 1 violent crime, but maybe now it's 4% likely to turn into a homicide instead of 2%.
All this said .. could you answer my original question? "(Counter) citation needed?" I'm getting the impression that you're so confident of "an increase in violence" over these decades that no new evidence will change your mind, and I'd really like to know whether the explanation is that there's some far-more-compelling old evidence that you've neglected to mention, or whether this is just confidence not based on evidence. I can come up with a dozen reasons the latter sort of confidence might exist (witness the long tails of these responses - surely the news wouldn't hammer on a category of story 24/7 if it was about as common as deaths by lightning!) but I'm hoping to stick with the former for myself.
roystgnr
11mo ago
"radiation" - not the real kind, like what supposedly comes from power lines and cell phones and such.
Real radiation does come from power lines and cell phones and such; but it's just low-frequency low-intensity EM, not ionizing radiation. And IMHO it wasn't crazy to fear possible health impacts anyway, a priori, so I'm glad we keep looking for a cancer link. It just seems like the harder we look the less chance of a link (and the lower the likely maximum effect size) we find.
roystgnr
11mo ago
what little there is, isn't near I-35.
It crosses I-35, twice. It's not a perfect substitute for I-35 because it doesn't parallel it; the "Red Line" (optimistically named to be forward-compatible with future dreams of having a second color too...) connects downtown to the NW, whereas I-35 is still required for anyone connecting to N, NE, S, or SE. If you want to use rail to skip (most) southbound I-35 traffic you have to divert 3 miles west to the Howard Park&Ride first.
Austin's population is about a million.
2.5M, if you include the whole metro area. And the metro area is very affected by these decisions; a ton of I-35 traffic is commuting from Pflugerville or Round Rock, and the Red Line goes through Cedar Park and Leander. $7,500 per resident goes down to $3,000 per metro resident.
some of the people impacted don't live in the area
A lot of the people impacted don't live in the area - if you want to go between Dallas/Ft-Worth (6.5M people) and San Antonio (2.5M), you either take a toll loop (expensive for car drivers, speed limits too high for many truck drivers, miles of extra distance) or you slog through I-35. The relative amount of impact is surprisingly small, though, with something like 85% of I-35 traffic from within the metro area.
one small part of the city.
The plans I've seen are divided into an 8 mile stretch in the center of the city (the expensive part), 8 in the south, 11.5 in the north. That's almost the entire North-South length of the city! And because Austin's historical philosophy toward East-West arterials has been "What's an East-West arterial?", a lot of travel which isn't really North/South as the crow flies gets fed into I-35 for a congested stretch anyway.
I also wouldn't chalk the whole cost up to "cars are heavily subsidized". The expensive new lanes are slated to be HOV-only, in part to make buses more attractive by no longer forcing them to sit in traffic with single-occupancy commuters. Some of the new features are things like decks and pedestrian bridges, connections between bike paths, etc.
On the other hand, I wouldn't bet on $7.5M being the whole cost or on everything planned being a completed benefit. The $7.1B "Project Connect" expansions to public transit (which you might interpret as "non-car-users are heavily subsidized", to be fair?) have been downscaled to a useless shadow of what was originally promised to voters.
roystgnr
11mo ago
the go-to political claim by the respectable institutions is that crime is caused by poverty.
Put scare quotes around "respectable" next time. And yeah, poverty only even correlates with a fraction of the problem.
We have a homicide rate that's increased or kept the same
And you say "since 1930s"? No. The 2020 jump leaves us worse than 1937-1939, but it's still below the start of the 30s and nearly 20% below the peak. The first big jump in homicide was over the 1900s through 1920s (following a long secular decline), and then the mid 30s through mid 50s was a decline again.
But though since the 1930s the US homicide rate fell a little again on net, the "huge jump from 1960 to 1980 then decline again from 1980 to 2000 then sudden more moderate jump in 2020" pattern is more complicated than that. This roller coaster is an interesting phenomenon but you have to pay attention to the details, not oversimplify. "We screwed up something horribly between 1930 and today" would have us looking in the wrong places, if the problem is really that we screwed up something super horribly between 1910 and 1930 and then again between 1960 and 1980 (or between 1890 and 1910 and then again between 1940 and 1960, if the "childhood lead exposure" theories are right) and we've fixed something between 1990 and 2010 but only part way.
If anything, we still could use way more details. E.g. I'd love to see that "murder correlates astonishingly well with single parenthood rates" graph extended in time instead of just space; looking at just national data they did increase together but then when the homicide rate fell the single parenthood rate didn't.
Why, on the basis of this data would we conclude there's less violence ?
As another comment here just paraphrased today: "if someone is biased towards something, then when presented with evidence that reinforces the bias, they think "CAN I believe this," but when presented with evidence that counters the bias, they think "MUST I believe this?"" This is not a straight path to truth.
The other data I've brought directly concerns the violent crime rate rather than trying to extrapolate from a biased subcategory of it. Ceteris paribus, far fewer people admitting to having been victimized is evidence of fewer victims! You've come up with the possibility that the ratio of crime to reported crime (and the ratio of crime to surveyed crime!? the ratios of surveyed to reported crime haven't changed too much) both increased a lot, not because you've brought evidence of that but because that would let you answer "CAN I believe this is wrong" in the affirmative. You're simultaneously neglecting the possibility that the ratio of (attempted) homicide to other violent crime increased slightly, because that is necessary to let you answer "MUST I believe this is right" in the negative. If now 4% of violence ends in death instead of 2%, like the data seems to show, that would have interesting implications ... but if your priors are "violence is simply proportional to homicide" then "true" is no longer a conclusion you can reach, it's in a blind spot that gets filled in from assumptions instead of evidence.
roystgnr
11mo ago
don’t see the downside of restricting one of the most useful technologies we’ve ever created
Like missing out on "... a Mars visit, and also a grand unified theory of physics, and a proof of the Riemann Hypothesis, and a cure for obesity, and a cure for cancer, and a cure for aging, and a cure for stupidity ..."? ("The Power of Intelligence", Yudkowsky, 2007; now in video form!)
There's an important difference between "don't see the downside" and "see the downside, but also the upside, and concluded that the latter is larger". Even if their conclusion is wrong, the doomers are all very much in the second category. Nobody thinks superintelligence is some kind of evil magic that can never be harnessed for good; they just think that at this rate it's too unlikely to be.
I’m always surprised that folks in the AI doomer camp seem to be so tech positive, but
You know what they say about surprise - it's your brain's way of letting you know that something you believed wasn't so. In this case, I'd suggest "they're coming to conclusions based on affinity for general categories rather than analysis of specific distinctions" might be the belief to ditch.
I personally think restrictions would do more harm than good, though. We'll get to AGI eventually regardless, and the more hardware overhang that's built up when we get there, the less crazy a rapid "foom" scenario looks. Our best odds now aren't to get the whole world to coordinate until we have proven safety via mathematical theory without experiments, but rather to hammer on safety as we improve capabilities and hope our results extrapolate to superintelligences too. "Hope our results extrapolate" might be in vain, but not so certainly as "get the whole world to coordinate" or "proven safety via mathematical theory without experiments".
They're not; you want what Google is calling a "TPU" and what NVidia is calling a "Tensor Core GPU" - operations on ridiculously coarse data types at ridiculously high speeds. Science+engineering simulations want FLOPS on 64-bit numbers, and video games want 32-bit, but AI is happy with 8-bit and doesn't even seem picky about whether you use 4 or 5 or 7 bits of that for mantissa.
I'd guess a cap on FLOPs (well, OPs, on whatever datatype) and another on memory bandwidth would work for the current software paradigm, for "serial" (as much as you can call a chip with 64k multipliers "serial") runs ... except that neural nets parallelize really well, and there's probably still a lot of room to improve interconnect bandwidth and latency, and if you do that well enough then at some point you don't care so much if there's a cap on serial execution speed. Human latency is enormous; no need to beat it by too many orders of magnitude.
The depressing bit is that the "hardware" side of the pair might be "just reuse the existing hardware with this new super efficient software". Even if the initial cap is low enough that we can't get to an AI smart enough to "foom" itself, if there are clever and heretofore unseen software improvements possible (and that's the safe way to bet) then human researchers will hit on them themselves eventually.
More options
Context Copy link