Posts
Comments
Those redditors have pretty weak arguments. The first comment is basically "the other academics all agree with the popular claim that Gilley is criticizing, so the popular claim must be true". The second guy basically states "Gilley correctly argues that Hoschild's evidence for a population decline is too weak. But if the evidence is bad, Gilley can't prove there was a genocide. Therefore Gilley is wrong".
The King Leopold thing is fake by the way. Bueno de Mesquita's account is based on "King Leopold's Ghost" which is a work of dishonest scholarship. Basically Hoschild used selective quotations and an intense blindness to the context to frame Leopold's adventure's in Africa as exploitative when they were really altruistic. For example, the hand-cutting quote "I will have to cut off the hand of every villager to meet my quota" is a cut from a longer passage that says "we have to adjust downward these quotas because the current demands on my unit are unworkable. To actually implement this I will have to cut off the hand of every villager". Even the policy entrepreneur's he quotes to make them seem anti-leopold were writing essays with the thesis "The Belgian's here are under resourced which creates problems, so we need greater Belgian involvement". Source https://www.theamericanconservative.com/king-hochschilds-hoax/ So I think the importance of political culture and ideology is understated in the opening anecdote.
In general the model does make a good starting position.
Thanks this was really helpful!
That was a typo, there was originally a note explaining that the line is sarcasm.
Thanks I'm really glad you identified that before it went elsewhere to create problems.
"Once you know which side you’re on, you must support all arguments of that side, and attack all arguments that appear to favor the enemy side; otherwise it’s like stabbing your soldiers in the back—providing aid and comfort to the enemy. " - Big Yud
I really doubt that people in the comment section will start siding with the RSF or the SAF and turning arguments into soldiers here. To almost all westerners Sudan todays is as distant as "Louis XVI during the French Revolution".
It's a good question. The obvious answer is that I was not comparing the current conflict with the most deadly conflicts since WW2, but with typical conflicts in Sudan and its environs over the past 20 years. I would guess Sudan has seen maybe 15k direct deaths so far (I believe official figures undercount Darfur and are far too low). Indirect deaths will be much higher due to the economic devastation from the war reaching Sudan's capital. The conflict shows no signs of stopping, so we can expect the ultimate death toll to be higher.
I don't have time to grab data, but only a small minority of conflicts enter the tens of thousands of battle deaths (which are easier to measure than civilian). By Wikipedia there are 62 conflicts since 1950 over 25k deaths. Total conflicts with 100 or more battle deaths is easily in the thousands, most of which you have never heard of like the Heglig War.
I expect that the death rate per month will decline over the course of the war until peace happens, at least in Khartoum and central Sudan. I don't understand the Darfur dynamics that produce massacres there so I cannot speculate on that.
I already have upvotes! Huzzah!
For new readers: SSA = Self-sampling assumption. Read that in Bostrums "Anthropic Bias". SIA might mean "Sampling independence assumption" but I am just guessing.
Does anyone have a good piece on hedging investments for AI risk? Would love a read, thanks!
Hahahaha that sounds like the worst value for money intervention I could possible do to become sexier. I've heard the surgery is super painful and debilitating when you do it.
This is helpful. I started taking creatine but got lazy about it, I'll get back on it.
As far as strength training, I started getting great female attention before I put on much muscle. I've become much more time constrained because I work like 55 hours a week anyway, so I only work out once or twice a week. Thanks for the recomendation on the youtube channel.
Exactly, this is the ring everyone else is optimizing for. So it’s tough to get relative to the other interventions.
Bumble, Hinge and Tinder.
I averaged that last time I was single. Should be able to get back there.
There is a failure mode here of overinvesting in status signals and underinvesting in being a pillar of your friend group.
I already have a good "status" so it's not a priority anyway, relative to the other areas.
That's helpful, thank you.
Do you know a trustworthy and concise source about how to Keto? The time to find a non-terrible guide via google sucks.
Haha yeah status is sexy!
The main reason is just that status is ambiguous between a "trait" and a "proof". Status is attractive partly because it mentally healthy, socially intelligent men will rise in status faster. But there's also an element of status being intrinsically useful because it's a resource to provide for a family.
The most efficient status-increasing interventions are all about presentation. Like I could get a white-house job to increase my status, but that would be super hard work. Earning the respect of my friends and advertising my career successes would also increase my status and is way easier. So I'll address it in the "proofs" post.
This an interesting essay and seems compelling to me. Because I am insufferable, I will pick the world's smallest nit.
The Wright Brothers took 4 years to build their first successful prototype. It took another 23 years for the first mass manufactured airplane to appear, for a total of 27 years of R&D.
That's true but artisanal airplanes were produced in the hundreds of thousands before mass manufacture. 200k airplanes served in WW1 just 15 years in. So call it 15 years of R&D.
Apologies if this has been said, but the reading level of this essay is stunningly high. I've read rationality A-Z and I can barely follow passages. For example
This happens in practice in real life, it is what happened in the only case we know about, and it seems to me that there are deep theoretical reasons to expect it to happen again: the first semi-outer-aligned solutions found, in the search ordering of a real-world bounded optimization process, are not inner-aligned solutions. This is sufficient on its own, even ignoring many other items on this list, to trash entire categories of naive alignment proposals which assume that if you optimize a bunch on a loss function calculated using some simple concept, you get perfect inner alignment on that concept.
I think Yud means here is our genes had a base objective of reproducing themselves. The genes wanted their humans to make babies which were also reproductively fit. But "real-world bounded optimization process" produced humans that sought different things, like sexual pleasure and food and alliances with powerful peers. In the early environment that worked because sex lead to babies and food lead to healthy babies and alliances lead to protection for the babies. But once we built civilization we started having sex with birth control as an end in itself, even letting it distract us from the baby-making objectives. So the genes had this goal but the mesa-optimizer (humans) was only aligned in one environment. When the environment changed it lost alignment. We can expect the same to happen to our AI.
Okay, I think I get it. But there are so few people on the planet that can parse this passage.
Has someone written a more accessible version of this yet?
Apologies if this has been said, but the reading level of this essay is stunningly high. I've read rationality A-Z and I can barely follow passages. For example
This happens in practice in real life, it is what happened in the only case we know about, and it seems to me that there are deep theoretical reasons to expect it to happen again: the first semi-outer-aligned solutions found, in the search ordering of a real-world bounded optimization process, are not inner-aligned solutions. This is sufficient on its own, even ignoring many other items on this list, to trash entire categories of naive alignment proposals which assume that if you optimize a bunch on a loss function calculated using some simple concept, you get perfect inner alignment on that concept.
I think Yud means here is our genes had a base objective of reproducing themselves. The genes wanted their humans to make babies which were also reproductively fit. But "real-world bounded optimization process" produced humans that sought different things, like sexual pleasure and food and alliances with powerful peers. In the early environment that worked because sex lead to babies and food lead to healthy babies and alliances lead to protection for the babies. But once we built civilization we started having sex with birth control as an end in itself, even letting it distract us from the baby-making objectives. So the genes had this goal but the mesa-optimizer (humans) was only aligned in one environment. When the environment changed it lost alignment. We can expect the same to happen to our AI.
Okay, I think I get it. But there are so few people on the planet that can parse this passage.
Has someone written a more accessible version of this yet?
Okay, let's do that backwards planning exercise.
In the long run, I want to do my research but live a low stress and financially comfortable lifestyle. The traditional academic path won't achieve that because I will end up doing my research but leading a high-stress and financially fraught lifestyle. There are three possible solutions to the problem, in rough order of preference A Pick a research agenda that is lucrative, so that I can supplement my income with lucrative consulting gigs and have a strong exit option B Learn to code and get a data science job, then do my research as a hobby C Get a government job related to my field (intelligence or aid)
Path A seems like the best one for both personal and EA reasons. Right now I split my time between writing on foreign investment and cabinet formation. But only the foreign investment work might pay the bills, the cabinet work ends with me in the brutal academia rat race. However, the foreign investment research might or might not succeed depending on contextual factors like competition, my ability to build a brand and the value of academic prestige in the field. So I should first try and figure out if the investment-academia path is satisfying.
I want to find out if that works over the next 6 months or so while in my academic program.
If the returns are too small and the competition too stressful, I should pivot toward a programming career. It's a well-payed 40-hour industry, and I can do my research as a hobby for 8 hours a week. That sounds like a lovely life too. So if I pick that, I would deemphasize my research and focus on coding skills for interviews and building career capital there.
I'm satisfied with that plan. The next question is, how do I stick to it? More on this later.
Just got Jason Brennan's book. It's very helpful!
That's a good question Barry.
Yes I could do a 3 paper very easily. I just finished a first article on expropriation and successions crises, it has a shot for a top journal. I'm working on a next one on succession crises and appointments. My professors tend to say that this isn't enough, that I need a special incredible dissertation where everything is laser focused on one topic and tightly linked. They also say that 90% of students take more than 5 years. I'm honestly confused.
Thanks for sending the link. I go to Dr. Brennan's school, so I can read the book then talk to him. Good idea!
| They're almost as horrified as people who've tweeted for years about sex and astrology and pineal glands are to discover that half their mutuals are actually LessWrongers.
I cracked up at that
Thanks! An error in my markdown was causing most paragraph breaks no to appear. Fixed.
Jesus christ. If I made that kind of money I could literally retire in a decade and then do whatever I want
I learned to code in R pretty well during my PhD, and I do enjoy it. It's usually relaxing, solving the problem feels good when you get it. I'm better than my colleagues at debugging and problem solving our code (data engineering mainly)
To be clear, you are talking about the salary for software engineers. Is that a better ladder than data scientists or data engineers? (my skills are closer to either of those fields currently)
I did quite a bit of research on it after this. It turns out there really isn't good data, the best is from the APSA but is full of holes. I did a tweet thread on it a while back.
I do have more publications than my competitors. Unfortunately, I have been repeatedly told in my program that publications do not matter and only dissertations matter. Kind of sucks, but what can you do. Publishing is definitely a signal of value, so I have the skills to do a good dissertation. It just sucks that what I like doing (papers) isn't rewarded.
The real kicker here is that even if I get the tenure track job, it's just not that great. For tenure track the average pay is 75k (for non-tenure 60k). More importantly, the tenure process is 6-8 years and very stressful. So I would be on the treadmill of competition from 27 (now) to 38. I doubt I want that level of stress for that long.
So probably not my best option but we'll see.
Also it's not that bad. I just finished a masters for free, I learned the classic causal inference methods. I can apply for sweet government jobs, government consulting, or learn to code.
Honestly, I wouldn't choose being a professor over my other options even if I could skip there right now. The low salary and location suck. I feel kind of stupid for not realizing this earlier, but I was idealistic at the start.
Going into academia was a mistake. It takes years of sacrifice and lots of luck to become a professor. The optimization is so intense you actually have less control over your research than you think. But even worse, being a professor sucks (in poli sci at least). You probably have to move to a rural area, the pay is like 60 or 75 if tenure track. The hours are the same as a normal job. The only benefit is doing your own research, but the pressure to compete squeezes the fun out of that.
I think someone wrote classic LW post about this. Yud mentions it in Inadequate Equilibria. Anyone know where that is?
Lol, and they say academia is a cult!
I was teaching my students Huntington’s clash of civilizations last week, an essay with similar problems. I had them nail down the testable assumptions, causal arguments, and falsifiable predictions of the piece. Got them to emotive the fuzziness themselves. It was a pretty rewarding way to teach.
Haha agreed.
I totally agree that "autocracy is always and everywhere an expectation phenomenon". My favorite piece of evidence is how quickly regimes collapse when the leader is terminally ill. Nothing has changed but you found out the Shah has cancer so you immediately throw down your arms. Because "Hello prince, i killed people for your dad now rob the people to pay me" doesn't work. Clearly, repression is motivated by the expectation the incumbent will win and pay you back in the future.
Yes, if people expect democracy to fail it probably will. But the inverse is not true. People expecting democracy to succeed is not nearly a sufficient condition for its success, and such expectations are more common than successful democratizations. The Russians really expected to democratize in 1992, and their experiment failed. The French really expected to democratize in 1789 and didn't. The Ethiopians I talk to today really expect Ethiopia to stay democratic and it obviously won't.
The US didn't just believe in themselves and win the gun game. They denied coercive capacity to the president and distributed it among state governors. They then constrained the governors with the threat of tariffs to prevent secession or shirking. The governors are specialized leader-restraining elites with coordination capacity, the ability to punish each other for shirking, and they are competitively selected.
The Chilean regime had strong expectations of democratic continuity but a terrible constitution that gave Allende the presidency with 35% of the popular vote, leading to collapse into autocracy.
On the other hand, society wide expectation flips are surprsingly common. I agree that it's weird, but it's true. Autocratic regimes (not leaders) are very shot lived. The oldest autocratic regime today is Saudi Arabia, which became a state around 1920. Even Saudi is currently in a massive consolidation crisis. The CCP is ancient at ~80 years old, and also in a consolidation crisis. Consolidation means the leader is systematically removing competent elites to cement control. When a consolidated leader dies regimes often collapse. Most autocratic states have not had a regime last 30 years. So it seems like the autocats expectation equilibrium would be very stable, but empirically it is quite unstable.
One could argue that the expectation that any regime will keep power is weaker than the expectation that "democracy will backslide into autocracy". I think that's a stretch, but this post is already too long.
I have been doing political betting for a few months and informally compared my success with strategies 1 and 2.
Ex. Predicting the Iranian election
-
I write down the 10 most important iranian political actors (Khameini, Mojtaza, Raisi, a few opposition leaders, the IRGC commanders). I find a public statement about their prefered outcome, and I estimate their power and salience. So Khameini would be preference = leans Raisi, power = 100, salience = 40. Rouhani would be preference = strong Hemmeti, power = 30, salience = 100. Then I find the weighted average position. It's a bit more complicated because I have to linearize preferences, but yeah.
-
The two strat is to predict repeated past events. The opposition has one the last three contested elections in surprise victories, so predict the same outcome.
I have found 2 is actually pretty bad. Guess I'm an expert tho.
Yes! That’s exactly what I was suggesting! Couldn’t have put it better myself.
I guess the full picture is some kind of co-evolution of institutions and popular opinion. Institutions channel human ambitions into behavior. Humans can uphold the institutions, or dismantle them, or pervert their intended function from inside. Maybe we need to wait 10 years after the institution was established, to see whether it works as intended.
I think you’re right, non-elite support for democracy is essential. I think elites are status maximizing assholes always and everywhere.
Problem is, no matter what kind of mechanism you set up, it only has a chance to work if a large number of participants are non-assholes. A bad moderator may e.g. censor his political opponents, and you may set up some system where users check his behavior, but what if most users agree actually support that?
I disagree what happened on the US after independence. The founding fathers were assholes, bad moderators who sought illicit advantages. They were checked by governors, voters and legislatures.
In other words, you can design a system where every actor pursues their own interests (is an asshole) but it doesn’t revolve into dictatorship. A longer treatment https://academic.oup.com/qje/article-abstract/126/4/1661/1923169?redirectedFrom=fulltext
For a very simple illustration, imagine you are in a room with 6 others. 6 of you have dollars, and one has a gun with one bullet. There is a Nash equilibrium where each of you give up your dollar and a NEwhere you say “shoot one of us am we’ll kill you”. Both only involve people being “assholes” or selfish. The point is to design a system where being selfish leads to good governance.
Misc. If and when median voters support genocide is a separate question from democracy. https://www.econlib.org/archives/2011/02/reflections_on_6.html Separation of powers is surprisingly bad https://www.bu.edu/sthacker/files/2012/01/Are-Parliamentary-Systems-Better.pdf
Slovakia wound up a flawed democracy or an anocracy, yes. Next door the Czech Republic wound up a full democracy with redistribution.
Yes there are other factors, policy does not explain 100% of political survival. And charisma does help one win popularity contests. You could write a post about it.
GJM’s interpretation is my intended meaning. Specifically I meant Pelosi’s intransigence, Boris’s strategic position changes and fake awkwardness. For khameini I was referencing the nuclear bullshit which is destroying Iran’s economy and obviously a bad deal, but good for his career. As evidence it’s a bad deal, observe that only 1 other country trolls the big five with nukes. But all of this allowed Nancy to stay speaker for years, Boris to go from mayor of London to PM and khameini to rule for 30 years
I love this.
Those are some really strong critiques. The framework did do something valuable for me. I have a few professors at my PhD program who are properly clueless. I've been trying to speak straight talk to them for a while, with negative results. It just strains the relationship. After reading this, I will try some babytalk. Frame my research agenda with some woke jargon, stuff like that.
Also the passage on woke talk and professors is spot on.
Great! Now redo it with equations included ;)
If the school shuts down the kids will just go back to the street. We do not send kids back into school when we observe transmission from kids being out of school. The evidence from Emily Oster suggest that there isn't much difference in transmission.
Also, I would argue that a small amount of transmission is worth educating our children, especially with 70-80% of the vulnerable vaccinated. Overall dividing life years lost by transmissions comes to 2 weeks per confirmed infections, so call that the base cost. Reduce it by 75% for targeted vaccination and each case is costing ~3 days of a persons life. And the student infections are the least dangerous kind. I could go either way on it if the alternative were no transmission. Since the alternative is about the same transmission rate but somewhere else, I say keep the schools open.
OTOH, the incentive argument is much stronger. Maybe the collective punishment forces the school to internalize the cost of transmission, leading to a pareto improving safe-school equilibrium.
I hope to find enough time to address this later. The foreign actors are affecting the revolution in two days. The western powers have revoked all aid and trade privileges, damaging the economy. The regional actors tend to side with the expected winner. The internal actors then update off the foreigners expectations.
Possibly the incentives on the parties are more important than the incentives on the individual candidates. We should then see a difference in issue-position flexibility between prop rep and single-member-district systems.
this is good and you should feel good
The two bottom predictions have already resolved. Large protests did not end and greater than 20 protestors have been killed so far.
Is there a clear resource about how Zvi formats and scores his weekly predictions?
Very interesting! I'll keep watching.
Thank you! More is coming :)
The most likely is a military challenger unseating Hlaing or the military's own party overthrowing them.
There are a couple of ways out. There's an unusual cohesion in the military currently, which allows the military to pull this off. Normally military regimes are unstable because even a small faction can threaten a civil war and force a regime change. So if the current generation dies -or- becomes dependent on their intelligence agency -or- a new officer faction things change. The new faction may prefer a return to the barracks, and change the whole system.
The western sanctions do not matter. Western investment, aid and loan forgiveness do matter, but no enough to stop the violence.
Speaking of public pressure to adopt better policies, let's form a twitter campaign to #unclogthefda. We're campaigning to decrease FDA red tape and accelerate vaccination approvals using tired-and-tested healthcare reform organizing techniques! You can read and comment on the plan here https://www.lesswrong.com/posts/QYkMWMZqQg49SrTdf/unclogthefda-a-twitter-storm-to-approve-vaccines