mycologos

Any comments section would require moderation, and internet moderation is a huge pain in the butt. It's not an unsolvable problem, as the examples you give demonstrate, but it does require a lot of effort, and I think you'd be pretty hard-pressed to find academics willing to expend that effort.

Yup. I don't mean "here's how you can use my theorem to build a bridge", I mean "here's an instantiation used to prove a more tangible result". Bonus points if multiple such results fall out of the general theorem. That's good evidence to me that generality is actually accomplishing something.

The target audience point is fair. There are basic concepts that I don't bother defining in any paper I write, and I'm sure there are people who would make the same argument about them that I'm making here. But I'm not suggesting writing everything like "An Extremely Gentle And Slow Introduction to X, With Lots of Reassurances That You Can Do It".

The question probably comes down to how accessible a paper should be. Personally, I think a reasonable bar is something like: a third-year PhD student in your broad area should be able to skim the paper and say a couple of paragraphs about what's happening and why it matters, and upon reading the paper more closely, present it in a seminar and defend it at least a couple of questions deep. IMO, most papers are not at this level, and are instead pitched at actual experts.

I think my point (and TFA's point as well) is that going from an experts-only paper to a seminar-ready one is actually not a ton of text. It might increase a paper's length by 5%. It's not about making everything longer, but adding enough context and signposting that the story can be followed at multiple levels, from the 5-minute pitch you'd get at a poster session to the every-detail one the author has. So I think we can have a paper that both the grad students and experts like.

I think doing this takes some skill and effort, which is well within the reach of most of the people who can write these papers, but way less effort is expended on this non-technical aspect.

As somebody who works in a mathy subarea of computer science, oh man, I agree. My heart always falls when I need a result and it turns out the original paper is some terse typewritten notice from the 70s whose first sentence is a definition with a bunch of proper nouns and whose main theorem is given at the most general possible level with no applications at all.

I have talked with math people about why this is, and responses are some combination of

a) being concise and being elegant are the same, same for maximum generality/abstraction

b) the people who should read the paper don't need things explained

c) I am afraid that some smart egotistical professor whose opinion I value for some reason will call me soft if I add extra handholding material

(Nobody has ever really said c, but my sense is it's true. Academic writing has a lot of imitation of style to prove you're part of the in-group.)

I think the economic returns to "decent understanding of X plus decent communication skills" are much higher when X is math than when X is art or language or history, so you need a greater passion for teaching to select it in spite of that fact, and this shrinks the pool of good math teachers relative to other subjects.

> I want nobody to have more than two children.

> It is ok for the population to shrink a little.

The first idea is way more extreme than the second idea.

A high school education doesn't go as far as it used to, women have more life paths that don't involve being a stay-at-home mom, houses are harder to come by, average age at first marriage is almost a decade higher than it was in the 50s ... notable, sure, but interesting, I dunno.

Looking at their SEC filing [1], the CEO compensation breaks down into ~$1mil of cash and ~$192mil roughly split between stock and options. Both the stock and options appear to have further conditions attached to them that reduce that number further.

I'm not even a fake accountant, so I don't understand the exact nature of the dilution, but the $193mil overall number seems to assume a lot of stuff goes extremely well in the IPO. I think the stock number assumes a stock price of ~$33.

[1] https://www.sec.gov/Archives/edgar/data/1713445/000162828024...

The insistence in both of your comments that the people at Boinc are not only taking advantage of people but "laugh[ing] at them" is odd. Why must your boogeyman be as unlikeable as possible?

Ah, anonymous was the wrong word. I meant instead that the shared data tries to restrict itself to information that doesn't obviously fall under a right to privacy. For example, trip times are shared, but locations are not.

My parents got one 15 years ago and it's still in great condition with no real effort on their part. I have no idea what has changed at Ekornes in that timespan though, for all I know they got acquired by private equity and turned to crap.

I read a book about chairs a few months ago, and one bit from it is

> Eighteenth-century furniture was expensive. At a time when a journeyman joiner earned three livres a day, a good-quality armchair might cost as much as a hundred livres.

I think there are a few reasons to be mad about this, but some are better than others:

1) People don't understand they're being monitored. I think this is a good reason to be mad. People should have some understanding of the agreements they make. It's part of being a functional adult in the world. It's also annoying that the companies keep spinning this as a tool to improve your driving, when it's clearly an attempt to price insurance against a person's actual driving habits.

2) The system's assessments are opaque. I don't have a good sense of how accurate any of these measurements are, nor what system is in place to ensure that. If the information collected is consequential enough to double a person's insurance costs, there should be some effort expended to be confident that the collected metrics actually reflect reality. I didn't see anything like that in the article, maybe I missed it, but it shouldn't just be some random team in a private company doing their best.

3) People's driving habits shouldn't be shared with insurance companies. This one ... this one I think is not great. It looks like the shared data at least tries to be anonymous -- they share driving behavior and times, but not actual location data. Heck, I'd be fine with scrubbing the times and just sharing the hard start and stop and speeding numbers (assuming point 2 above is addressed). I get that a knee-jerk defensiveness about privacy would make Thomas Jefferson proud or whatever, but we strike balances on public welfare and private freedom all the time. If you're itching to manufacture 1 gram of ricin to put in a sealed glass vial above your mantle, too bad, you can't. Cars aren't ricin, but they are the shortest path between most humans and homicide. If this kind of intervention induces people to be more careful to keep their current insurance rates, I think that's reasonable. Driving like a maniac is not a human right or a protected characteristic.

If it's so telling then tell us, enough with the dark innuendos!

I think this is an underrated point in these discussions. There are three components to every job: how much value other people get out of you doing it, how much value you get out of doing it, and how many other people want to do it. Whether it's game development or adjunct prof'ing, employers know when the job their offering has benefits beyond pure salary, and they adjust that salary accordingly. The humanities grad students I knew weren't stupid. They could see the statistics in front of them. But staying the university world for a bit longer, and seeing other people do the same, tempted them down that path, and continues to tempt a surprising number of them.

Where are people getting all the confidence they place in Sam Altman as a leader of a science/engineering effort? From my understanding, he has been a successful fundraiser and investor and that's kind of it. Those are real and valuable skills, but they don't seem so relevant to the task at hand. Or is the statement that Sam Altman will keep OpenAI funded (so somebody else can lead the work)?

At first I wasn't sure about your claim, but then you said it's obvious, and my doubts went away!

> and I'm skeptical that you're a moral human being

Hm, surely we don't need to go that far. This is an internet discussion, we're supposed to be nice to each other!

> Apparently poor people with families don't need larger vehicles the way middle class people with families do.

Maybe I don't understand what claim you're making: is it 1) vans/SUVs must not become any more expensive (via higher insurance requirements), or else poor people will be priced out of owning a car entirely, or 2) owning a van/SUV is almost a human right, and raising the barrier to it is immoral?

I've tried looking at listings and I don't see evidence of the claim that somehow poor people are forced to buy SUVs because they're cheaper. There appear to be many big and small cars at every price range.

For 2), huge cars have only been widespread for a decade or two. The idea that they're somehow necessary for a meaningful human life is hard for me to grok.

It seems like he's talked a lot about the T in LGBTQ, not so much the G? Evidence that he would be especially incensed at getting beat by a gay guy seems thin on the ground unless we insist on treating LGBTQ as a bloc.

(I don't exactly keep up with Musk's doings, though.)

There's an implicit claim here that it's easier to buy a cheap big used vehicle than a cheap small used one. That doesn't seem true, as even now at the height of SUV popularity, most new cars are not SUVs. A significant majority of used cars should therefore be small cars, no?

That's why I'm suggesting that we target very large vehicles first. I am skeptical that poor people absolutely need very large vehicles, which are more expensive to run anyway.

That's disappointing. I saw it was sold recently, but I was hoping it was one of those sales where things mostly continue as usual.

The Economist had an interesting recent article suggesting that, at least in America, part of the problem is that car insurance is way too cheap, so owning a car and driving like a moron is too easy: https://www.economist.com/united-states/2024/01/18/why-car-i....

> According to the Insurance Research Council (irc), an industry data group, 29% of claims nationally (and over 50% in several states) involve people insured at the state minimums [in the tens of thousands of dollars]. Few policies go beyond a few hundred thousand dollars of liability. The cost of a serious crash “is never going to be covered by that”, says Dale Porfilio, of the irc. By contrast, in Germany drivers are required to have €7.5m ($8.2m) of bodily-injury coverage, and in Britain liability is unlimited.

I don't have a car, but I acknowledge that most Americans live in places where not having a car would make their lives much harder. Still, it seems reasonable for me to raise insurance requirements on enormous SUVs. They're 50-100% more dangerous to pedestrians, so this would reflect a higher actual cost inflicted on other people. If you don't want to pay that cost, and you still need the transportation of a personal car, buy the smaller and less dangerous vehicle.

Tangential, but Serious Eats is one of my favorite food websites on the internet. In an ocean of crappy SEO recipe blogs overstuffed with generic photos and unnecessary backstories, Serious Eats has a deep collection of well-explained and high-quality free recipes (and articles like this one).

I would happily pay some kind of premium membership, but there doesn't appear to be such an option.

How useful is reducing to a semidefinite program in reality? A fair amount of stuff seems to conclude with "now that we've reduced to an SDP, it's all polytime from here baby, so we're done modulo boring implementation details that nobody cares about". But I've tried and failed to understand how meaningful that polytime is in a practical sense. Anybody know?

Prison labor is indeed a multi-billion dollar industry, but I question the idea that it drives "a lot of incarceration". For better or for worse, Americans seem to be pretty OK with how incarceration currently works: across almost every political or racial group, most people think the current amount of incarceration is either "too little" or "about right". In total, less than 1/3 of Americans think it's "too much" [1].

[1] https://www.pewresearch.org/short-reads/2021/12/06/u-s-publi...

I sometimes wonder whether sentencing people to time in prison is the right equilibrium. My understanding is that doing the stuff associated with long prison terms is also associated with not having the longest or most carefully considered time horizon. In that sense, long prison terms are a bad punishment because they're an insufficient deterrent that is very expensive, in terms of both state resources and the prisoners' time.

Unfortunately, any deterrent has to by its nature be unpleasant, so if prison terms are off the table, you end up with instinctively gross and backwards-seeming things like inflicting physical violence on prisoners. That, uh, doesn't seem good either.

For the sake of brevity, at the risk of glibness: it seems like a system created with little consideration for its users, who are very different people from its designers.

Diaspora is my favorite Egan book. Permutation City seems to get talked about more, but the dust theory stuff just felt implausible, and it has one of the worst sex scenes I've ever read. Maybe it's because Diaspora is less concerned with anything as abstract as consciousness and more interested in how different forms of life play out, which I find fun when it's done well.

> The prof I was working with told me that this anonymous reviewer was probably the writer of those papers and asking for citation.

I think this is mostly just a kind of fun way of complaining about reviewers. Having been a reviewer for many computer science conference, I've often had the names of other reviewers visible to me. Most requests for citations are not for the reviewer's own work, and truly irrelevant citation requests seem to be pretty rare.

I do perceive journal/lab-centric fields to be worse about this, though.

Note that the article specifies that these cartels mostly feature participants from bad universities in countries that have really whacked-out academic reward mechanisms that arose because some powerful central government force that doesn't understand research productivity came up with a dumb way to incentivize it.

Prestige is not a failsafe guarantor of quality, but it kind of is, in the sense that somebody with a reputation to protect is going to be more careful about coming across as "unserious", and doing stuff like this is a good way to get a bad reputation quickly.