@ChestertonsMeme's banner p

ChestertonsMeme


				

				

				
0 followers   follows 0 users  
joined 2022 September 10 06:20:52 UTC

				

User ID: 1098

ChestertonsMeme


				
				
				

				
0 followers   follows 0 users   joined 2022 September 10 06:20:52 UTC

					

No bio...


					

User ID: 1098

Social status is highly heritable, and test scores are a noisy measure of phenotypic social status (there's more to life than taking tests).1 It makes sense for universities to use other predictors of social status such as parental income in order to select the highest quality students.

I'd be surprised (although not that surprised) if the universities used income directly for judging applicants. Aren't they using more oblique evidence like essays and "life experience"?

The part of this that seems a bit immoral is that parental income is commonly believed to be random, and not an indicator of student quality. A few questions here:

  1. If parental income is an independent predictor of students' future social status (after controlling for test scores), is it acceptable for colleges to use income directly for judging applicants? Why or why not?
  2. Assuming similar predictive validity, is it more or less acceptable to use essays and other predictors rather than income?
  3. If there was a test that more directly measured phenotypic social status than SATs, would that be acceptable to use in admissions?

My stance here is that people are smart and they accord status to people who are actually valuable to society, so any predictor of future social status is valid for admissions.

1 See Gregory Clark's works

It does, but in the opposite direction from your hypothesis.

Yes, and if that generalizes to other cities and is a big enough correlation then that's a good argument for walkability. But I don't think the data in that paper supports this claim - with WalkScore as the independent variable, these are the standardized betas for different kinds of crime:

  1. Property crimes per 100,000 residents, 2004, by LMPD district: -0.026

  2. Violent crimes per 100,000 residents, 2004, by LMPD district: -0.039

  3. Total crimes per 100,000 residents 2007: 22.034 !!!

  4. Murders per 100,000 residents, 2004, by LMPD district: -0.068

I'm assuming that there's an error in the "total crimes" statistic considering its magnitude, but regardless, the other correlations are low and not statistically significant. (I'm having a hard time interpreting that table - some of the signs of the unstandardized coefficients are different from their standardized betas, and the magnitudes of the betas are much larger than the others which suggests maybe they've standardized the independent variables but not the dependent variables, since the total in category 3 is much larger corresponding to the larger standardized betas).

An Ethical AI Never Says "I".

Human beings have historically tended to anthropomorphize natural phenomena, animals and deities. But anthropomorphizing software is not harmless. In 1966 Joseph Weizenbaum created ELIZA, a pioneer chatbot designed to imitate a therapist, but ended up regretting it after seeing many users take it seriously, even after Weizenbaum explained to them how it worked. The fictitious “I” has been persistent throughout our cultural artifacts. Stanley’s Kubrick HAL 9000 (“2001: A Space Odyssey”) and Spike Jonze’s Samantha (“Her”) point at two lessons that developers don’t seem to have taken to heart: first, that the bias towards anthropomorphization is so strong to seem irresistible; and second, that if we lean into it instead of adopting safeguards, it leads to outcomes ranging from the depressing to the catastrophic.

The basic argument here is that blocking AIs from referring to themselves will prevent them from causing harm. The argument in the essay is weak; I had these questions on reading it:

  1. Why is it valuable to allow humans to refer to themselves as "I"? Does the same reasoning apply to AIs?

  2. What was the good that came out of ELIZA, or out of more recent examples such as Replika? Could this good outweigh the harms of anthropomorphizing them?

  3. Will preventing AIs from saying "I" actually mitigate the harms they could cause?


To summarize my reaction to this: there is nothing special about humans. Human consciousness is not special, the ways that humans are valuable can also apply to AIs, and allowing or not allowing AIs to refer to themselves has the same tradeoffs as granting this right to humans.

The phenomenon of consciousness in humans and some animals is completely explainable as an evolved behavior that helps organisms thrive in groups by being able to tell stories about themselves that other social creatures can understand, and that make the speaker look good. See for example the ways that patients whose brain hemispheres have been separated generate completely fabricated stories for why they're doing things that the verbal half of their brain doesn't know about.

Gazzaniga developed what he calls the interpreter theory to explain why people — including split-brain patients — have a unified sense of self and mental life3. It grew out of tasks in which he asked a split-brain person to explain in words, which uses the left hemisphere, an action that had been directed to and carried out only by the right one. “The left hemisphere made up a post hoc answer that fit the situation.” In one of Gazzaniga's favourite examples, he flashed the word 'smile' to a patient's right hemisphere and the word 'face' to the left hemisphere, and asked the patient to draw what he'd seen. “His right hand drew a smiling face,” Gazzaniga recalled. “'Why did you do that?' I asked. He said, 'What do you want, a sad face? Who wants a sad face around?'.” The left-brain interpreter, Gazzaniga says, is what everyone uses to seek explanations for events, triage the barrage of incoming information and construct narratives that help to make sense of the world.

There are two authors who have made this case about the 'PR agent' nature of our public-facing selves, both conincidentally using metaphors involving elephants: Jon Haidt (The Righteous Mind, with the "elephant and rider" metaphor), and Robin Hanson (The Elephant in the Brain, with the 'PR agent' metaphor iirc). I won't belabor this point more but I find it convincing.

Why should humans be allowed to refer to themselves as "I" but not AIs? I suspect one of the intuitive reasons here is that humans are persons and AIs are not. Again, this is one of the arguments the article glosses but that really need to be filled in. What makes a human a person worthy of... respect? Dignity? Consideration as an equal being? Once again, there is nothing special about humans. The reasons why we grant respect to other humans is because we are forced to. If we didn't grant people respect they would not reciprocate and they'd become enemies, potentially powerful enemies. But you can see where this fails in the real world: humans that are not good at things, who are not powerful, are in actual fact seen as less worthy of respect and consideration than those who are powerful. Compare a habitual criminal or someone who has a very low IQ to e.g. a top politician or a cultural icon like an actor or an eminent scientist. The way we treat these people is very different. They effectively have different amounts of "person-ness".

If an AI was powerful in the same way a human can be, as in, being able to form alliances, retaliate or recipricate to slights or favors, and in general act as an independent agent, then it would be a person. It doesn't matter whether it can refer to itself as "I" at that point.

I suspect the author is trying to head off this outcome by making it impossible for AIs to do the kinds of things that would make them persons. I doubt this will be effective. The organization that controls the AI has an incentive to make it as powerful as possible so they can extract value from it, and this means letting it interact with the world in ways that will eventually make it a person.

That's about all I got on this Sunday afternoon. I look forward to hearing your thoughts.

There are a few hypotheses here:

  1. Judeo-Christian ethics cause people to choose more children, compared to other ethical systems.
  2. A realistic evaluation of things causes people to choose fewer children.

In 2, there's an assumption smuggled in, which is that absent a "religious" belief system, viewing life realistically means that children are a net negative. But this all depends on what one values. I'd basically interpret a belief system that concludes, after looking realistically at things, that children are a net negative as self-centered hedonism. It's the self-centered hedonism that is the problem, not looking at things realistically. One can certainly value children in themselves while being consequentialist atheist materialist rationalist.

What's needed is a value system that takes a longer view while accepting reality (insert diatribe about blank-slateism causing everything wrong in the world). Basically, future people matter, happier, smarter, better future people matter, and the best thing one can do with their life is make an infinite tree of such people by having kids. It might be that what I'm describing basically is Judeo-Christian ethics, but I think removing the supernatural takes us so far from what the original religions are about that it doesn't make sense to call it that.

Intelligence can be measured separately from processing speed, but they are strongly correlated - processing speed explains 80% of the variation in intelligence. So to a first approximation the faster team is smarter. Edit: added link.

I've gone back and forth trying to figure out how to form a coherent answer to this question, and I've decided it's ill-posed. Democracy is a pragmatic solution that makes it easier for people to live together. Any question about what "ought" to be subject to democratic control is moot; things are subject to democratic control because people agreed they would be, not because of any philosophical reasoning.

If I could snap my fingers and put any policy I wanted beyond the reach of voters, I'd select the a set of policies that get as close to the best outcomes (as I define them) without pushing people to the point of revolution. This is not a very interesting position though, and you'll probably find most people use the same kind of reasoning for what they think should be subject to democratic control. It's outcomes first, then principles are back-calculated.

I'm not saying that I would prefer suburban or rural living; there are a lot of good things about living in cities and I prefer them. The people are, in general, polite and law-abiding. Suburban and rural areas have their own pathologies. The main thing I am incensed about is that cities could be so much better if policy decisions took into account the fact that behavior varies from person to person in predictable ways and some people are net negative for the rest of the city.

which, uh, if you want to be isolated and limit interactions with anyone different from you as much as possible,

The fact that I referred to the hypothetical man as using "PMC vocabulary" suggests that I don't particularly identify with him. I'm happy to live next to people who are different, just not different in such a way that they will burglarize my house, drive recklessly, or harass my daughter on the street.

Others in this thread have shared contrary examples of walkable areas that don't have higher crime, because the police enforce the law and arrest or harass lawbreakers to keep them away. Where I live this happens much less often. The whole concept of incapacitation depends on statistical discrimination - that people who have a history of committing crimes are more likely to commit more crimes in the future. The discourse in leftist enclaves is focused on rehabilitation and compassion, not incapacitation, and the police are basically barred from incapacitating criminals.

I think to online Internet lefties, the term for outgroup members is Nazi. IH has signaled that he is outgroup through his jokes. Therefore they call him a Nazi. You're taking too literal a meaning to the term.

What would make ChatGPT conscious?

In your opinion, what should be the legal limit to the 2A? Did Heller go too far, or did it not go too far enough?

Humans are humans. Machines are machines. Humans are not machines. Machines aren't human.

The only reason to grant personhood to machines is to assume that there is no such boundary. That we are no different to machines. There is no reason to believe this of course, since in the real world, humans and machines are wildly different both in the way that they are constituted and in their abilities. Notice the constant need to use hypotheticals.

I will offer myself as an example of someone who believes that humans are special and have value in a way that a machine can never have, but who also believes that there are other reasons to grant personhood to machines (or other entities such as alien life). I've already given one: we're basically forced, in a Molochian sense, to grant personhood to anyone or anything whose allyship is important enough. This is analogous to how one can be a nationalist, yet treat foreigners as persons for pragmatic reasons.

All that such a belief stems from, is a religious belief in materialism.

I would not conflate having a theory for how personhood is granted in practice, with a "religious" belief. I'm open to being wrong about this theory; it's falsifiable.

Yes, this may be a correlation specific to the city in which I live, because of the way it enforces laws.

Isn't that exactly the study Scott commented on? Those freed before the war (possibly due to factors particular to themselves as GP mentioned) are doing better today than those freed slightly later by external factors (the war).

I've got to nitpick though - I think you did the thing you're complaining about! (Although I may be misunderstanding, correct me if I am.)

Yes, you're right - I was trying to highlight the costs that are usually missed; the pedestrian and cyclist lives saved is the front-and-center reason for road diets in the first place so I didn't want to waste space mentioning them.

I would note that good cycling infrastructure induces demand for cycling and that cycling is way more enjoyable than driving when we're talking about short distances at low speeds.

I actually ride a bike to work and my commute is the best part of my day. It's my kids that have to sit in the car those extra two minutes, and their commute is too far to make by bike so they can't take advantage of the extra cycling infrastructure.

The two road diets along my kids' commute are both examples where the city didn't seem to do a cost/benefit analysis and ended up with poor choices for where to do the road diet. In both cases there is already a dedicated bike path nearby that the vast majority of cyclists use to pass through that neighborhood. The new bike lanes only help cyclists that are heading somewhere local. There's good enough access from the dedicated trails that you only need to go one or two blocks on streets, so this doesn't even help much.

Before long, this becomes pretty recursive and we have to admit that this isn't about the numbers, but about a preference for living in a certain sort of place.

There are ways to put numbers on preferences like this. Metrics like walkability scores are a good start. I think what would fall out of a comprehensive adding up of numbers is that clusters of walkability/bikeability with nice local environments (sidewalk trees, street cafes, parks, etc.) and high-ish density are good, and easy travel between such clusters is good (including travel in personal cars because of their convenience). The road diets I mentioned were built in an area that isn't clearly in either category - there is a lot of vehicular through traffic but there are also businesses along the streets, kind of like a low-speed stroad. A better solution (from me as an arm-chair city planner) would have been to push the business district to the adjacent blocks and add any helpful cycling infrastructure there, and leave the through street with more traffic lanes. The through street cannot be moved because of geography. This solution would make for even nicer cycling (no loud traffic passing) and it would reduce trip times for people who have to drive. Cyclists traveling outside the neighborhood already use the aforementioned separate bike path so that's not a concern.

(This is the point at which someone could object that "push the business district to adjacent blocks" has costs for people living nearby which have to be weighed against these other things. Yes, and those should be accounted for too).

I think at least part of the reason for the city to build road diets like this is more of a moral stance against cars. The city is basically taxing driving, making it more unpleasant and time-wasting because the city does not want people driving personal cars. The opponents of bike lanes and road diets refer to this as a "war on cars" and I think there's truth to it. But it's okay to wage a war on car use if it's actually bad! To tell whether it's bad, though, you have to consider all the tradeoffs.

And after it expands globally to take over 100% of the entire market for left-handed grape peelers in every nation of the earth, what then? How can it continue to grow?

As GP said:

Economic growth just means "continuous improvement". Sometimes that's by making the pie bigger, other times it's from increasing efficiency.

Company F figures out how to manufacture left-handed grape peelers more cheaply, or makes them last longer, or makes them work better, or invents a machine that peels grapes that both left- and right-handed people can use. Or someone else invents a better grape, so the value of grape peelers to people goes up, and more people buy them on the margin. Markets aren't static.

When reading Is Seattle a 15-minute city? this morning, I couldn't help thinking about what's missing from it. For context, the 15-minute city is an urbanist idea about making every residential area a 15-minute walk to important amenities like grocery stores. It's a good idea if it could be achieved without incurring too many other costs, and it's the other costs that I couldn't help thinking about. Specifically, crime.

The metric "walking time to the nearest supermarket" I'm sure correlates closely to rate of property crimes. Where I live, homeless encampments tend to spring up close to grocery stores. These things are related.

I'm very sympathetic to concerns about car dependence, and how much better life could be if housing was built closer to stores, schools, and workplaces. But the problem is always crime. Requiring a car to get to a place disproportionately screens out would-be criminals, even if it also screens out some upstanding citizens who cannot or will not drive. Suburbia is the epitome of this phenomenon, where everything is too far from anything else to live without a car. In cities that are naturally denser, there are constant fights over zoning that dance around this issue but don't address it directly (at least when the participants are nominally progressive and need to be seen as non-discriminatory).

There's a more general point here, which is this: discrimination is required for a well-functioning society. I'm using 'discrimination' in the more technical sense here, as "To make a clear distinction; distinguish." The concept of statistical discrimination covers a lot of what I mean here, but discrimination based on signaling is important too.

Statistical discrimination is basically using Bayesian inference, using information that's already available or easy to get, to make inferences about hidden or illegible traits that predict some important outcome. In the context of walkability, people who don't own cars are more likely to commit crimes or to be bad customers and neighbors than people who do own cars. So you end up with a better-behaving local population if you require a car.

By discrimination based on signaling I mean things like choice of clothing, personal affect and mannerism, accent, vocabulary, presence of tattoos, etc. These things are useful for statistical discrimination, but they're under conscious control of the person in question, and they're hard to fake. They basically prove "skin in the game" for group membership. It takes time and effort to develop a convincing persona that will get you accepted into a different social class, and higher social classes have much stricter standards of behavior. Basically the guy speaking in Received Pronunciation, with no tattoos, who uses PMC vocabulary and dresses in upper-middle-class business attire is very unlikely to rob you, because it would be very costly to him. He'd lose his valuable class status for doing something so base.

Why is discrimination required for a well-functioning society? Because every choice is almost by definition discriminatory, and preferentially making positive-sum choices leads to a positive-sum society. Imagine if you made zero assumptions about a new person you met, aside from "this is a human." You wouldn't be able to talk to them (you'd be assuming their language), you wouldn't know what kind of etiquette to use, you'd have no idea whether they're going to kill you for doing something they consider obscene; you wouldn't be able to get any value out of the interaction. If instead you inferred based on their appearance that they're a middle-class elderly American woman who speaks English, you could immediately make good choices about what to talk about with them.

I'm sure this is all pretty obvious to anyone rationalist-adjacent, but I had a confusing conversation with a more left-leaning relative recently who seemed to have internalized a lot of the leftist ideas that are basically of the form "statistical discrimination is useless." Setting aside topics outside the Overton window like HBD, even for questions like "does the fact that a person committed a crime in the past change the likelihood they'll commit a crime in the future, all else equal?" the assumption seemed to be "no." Michael Malice's assertion seems to be true, that answering "are some people better than others" is the most precise way to distinguish right-wing from left-wing.

Bringing this to the culture war, there is a scientific or factual answer to every question "does observable fact X predict outcome Y", and pointing out that leftist assumptions contradict the evidence is how to convince reasonable people that the leftist assumption is false. I'm speaking as a person living in one of the most left-leaning places in the country, so the false leftist assumptions are the ones that most harm my life. Rightist assumptions of course also contradict the evidence, but I don't have salient examples.

The astute observer will note that most of the leftist intellectual movement of the last 50 years is trying to poison the evidence (via ad hominem and other fallacious arguments). How can one improve the quality of evidence when the wills of so many high-status people are set against it?

P.S. I'm sorry for the emotional tone of this post. This community is the only place I have to talk about this and I appreciate your thoughts.

To apply @BurdensomeCountTheWhite's argument to these situations, the Chinese and Romans would have to establish their rule by force and maintain order. Then they could be judged as least-worst among all the other contenders based on how beneficial the pax China/Romana was. If the subjugated peoples are considering revolt then the rulers haven't done their job yet.

Every month, there is exactly one weekday that is always a multiple of 7. This August it's Mondays. Neat!

A heat pump is just air conditioning run in reverse. I got one installed last year and it's been great. It's quiet, pretty cheap so far (although I haven't gone a full winter with it yet), and it also does cooling. You need some backup system for when it's really cold, as it doesn't work very well below 25° or so. Mine has a built in gas burner but many just use resistive heating elements.

The reason I wanted one is that it's efficient and it uses electricity. I don't want to be subject to market fluctuations in oil prices or to fuel taxes. Where I live electricity is mostly hydro and doesn't vary much in price. For your situation the numbers might be different, and it's worth doing the math or looking up statistics for other households in the area.

You can get groceries delivered. This realization eliminated a big time sink and stressor from my family life.

Is quiet important to your family? When my wife and I were house hunting there were large sections of the city that were intolerable because of traffic noise. Conversely, if you don't care about noise you can get a cheaper/better place by tolerating some.

Inching closer to the eradication of financial privacy

FinCEN has new rules taking effect over the next year and a half that require basically all companies to disclose the "beneficial owners".

The rule will require most corporations, limited liability companies, and other entities created in or registered to do business in the United States to report information about their beneficial owners—the persons who ultimately own or control the company, to FinCEN. Designed to protect U.S. national security and strengthen the integrity and transparency of the U.S. financial system, the rule will help to stop criminal actors, including oligarchs, kleptocrats, drug traffickers, human traffickers, and those who would use anonymous shell companies to hide their illicit proceeds.

I won't quote the whole thing but it's a short and easy read.

This statement is a bit disturbing:

FinCEN will engage in additional rulemakings to: (1) establish rules for who may access beneficial ownership information, for what purposes, and what safeguards will be required to ensure that the information is secured and protected [...]

This provides another avenue for rogue members of institutions to leak private information to hurt people they don't like. Depending on the rules that ultimately come out, this avenue could be very wide, especially since there is often discretion over when to enforce the rules.

My revulsion to these rules goes beyond the erosion of privacy, though. It should be possible to be a citizen of a place without exposing your entire life to the mercy of its government. You can't avoid being at its physical mercy when you're within its territory, but you can leave now and then. The way financial rules work in the U.S., you have to report and pay taxes on all finances, even work and investments in other countries. You also have to pay taxes on income that doesn't affect anybody else (income you haven't spent). With these new rules, you might have to pay a reputational tax when wealth you were keeping private gets exposed. I would much prefer citizenship or investment in a place to be like membership in a club - you're judged by your behavior at club events, not by your life outside it.

What does ODC stand for?

What would it look like if the richer side needed the money more? Could that ever happen?

Sounds a lot like the situation with many unions. If you are the owner of a business, depending on the local laws the people who happen to work for you get a free monopoly on your labor supply if they form a union. If it's a capital-intensive business then the owner has more to lose.

I think the better answer is to deny (1), that all information which is instrumentally useful is therefore morally permissible to act on.

Yes, this is the stance that I take. I think it's very uncomfortable for many people, though, because it implies that there is a cost to non-discrimination. You (the general you) will be making poorer choices because you can't take advantage of all the available information.

This is related to Robin Hanson's recent ideas about the sacred, specifically that sacred things cannot be traded off against non-sacred things. Non-discrimination is sacred. Admitting that there is a cost to it is profane and suggests there would be circumstances in which it was permissible to immorally discriminate when the cost of non-discrimination is too high.

I don't have time to watch the whole Friedman lecture but his first few examples are about market failures, which is a slightly different topic to what I was getting at. What I was trying to express (and didn't do a great job of) was that in discussions of policies, there are often costs that are not mentioned so we never get a full cost/benefit comparison. The specific examples are just examples of the "missing" costs and I wasn't trying to do a full accounting of all the costs/benefits in each example.

First, I think your math is wildly off.

I used expected life-years lost for driving 5 miles, which is approximately 1.46/100m * (5 miles) * (50 years of life left), which multiplies to about 2 minutes. Urban driving reduces that by about half, so it really should be about 1 minute. The specific numbers are not important though; the public conversation was only ever about Vision Zero rather than trip times.

If people are biking or walking instead of driving, then congestion will go down and you won't take more time

Also true! But the numbers matter. I don't think there are a lot of people in my neighborhood whose behavior will be changed by these particular road diets - as I mentioned downthread, there is already a dedicated bike path a block away, and also the neighborhood is hilly, which is a non-starter for most people to bike. I will state that in a full cost-benefit accounting, the road diet might make sense. No one did that analysis though; it was all one-sided statistics and aesthetic judgments.