mordymoop

I have also heard them called that, but I agree it is an almost actively misleading label.

I didn’t find out until many years after graduating that the US is sprinkled with what are known as Primarily Undergraduate Institutions (PUIs), including a large number of well-regarded 4-year colleged which don’t have graduate programs, the professors don’t do (much) research (but do have PhDs generally) and the hiring and evaluation decisions are 90% based on teaching.

These institutions are usually on the small side and don’t make money in grants because they don’t do research, so they don’t spend millions in advertising or in sending out interviewers to thousands of high schools. So there’s a good chance you’ve never heard of even the ones you live near to.

These schools tend to produce both better experiences (less reliance on the painful “weed-out course” degree plan architecture of huge school) and somewhat better outcomes. I will almost certainly be sending my kids to be PUI and not a megaversity when they’re older.

It is also a lifesaver for complicated fiddly array index code. One of the most cognitively costly thing for me, decades into programming, is still anything related to array partitioning, or remapping indices. Copilot pretty quickly just gets what you’re trying to do and auto fills a correct line of code. It’s basically never wrong about this sort of thing. And even in the rare case where it’s not correct (usually because it’s mistaken about what I’m trying to do), it’s much faster for me to read the line and figure out what it does than it is to mentally slog through the problem and write the line from scratch.

The ramblings printed in the soap bottles have been the world’s longest running stealth ad campaign.

The purpose of sanctions is not to punish, but to apply pressure aimed at obtaining compliance, or at least compromise. The US doesn’t depend all that much on other countries for trade, and no other countries can orchestrate big international collective actions the way the US can. Imposing ineffective sanctions is, at best, a demonstration of impotence.

So there are two questions: who could have imposed meaningful sanctions on the US? The answer is probably nothing short of a concerted international effort amongst many nations. And then, the more important question, who would have imposed sanctions? Nobody, because the Iraq war had a substantial amount of international support, at least within Europe.

I’ve always found it amusing that when I was in school/grad school, I was reading a lot of material that I didn’t usually find interesting, so it would have been valuable to know how to take good notes. I never even heard of a course in the subject of note-taking so I never learned any of these methods you hear about. Nowadays I only read or otherwise consume stuff that I find interesting for its own sake so my retention is vastly better, and I have no real need for such systems anyway.

Or “charisma”!

Sure - I think we all learn tricks like that. But you learned that pattern of tokenization, it wasn't arbitrarily foisted on you.

What GPTs have to deal with is more like, you are fed an arithmetic problem via colored slips of paper, and you just have to remember that this particular shade of chartreuse means "224", which you happen to have memorized equals 7 * 32, etc., but then the next slip of paper is off-white which means "1", and now you have to mentally shift everything ...

When you toss “2241 + 19873 =” into an applet that shows you the default tokenization scheme GPT-3 uses, you get this:

(224)(1)( +)( 198)(73)( =)

I’ve heard it remarked before that, while tokenization is obviously an unavoidable part of a model with an architecture like GPT, this is a very silly way of tokenizing number strings for the purposes of learning or doing arithmetic. Indeed, I think a lot of GPT-3’s puzzling edge-case performance can be ascribed to weird and unhelpful tokenizations. Just imagine if you were forced to learn arithmetic with a brain that automatically categorized “224” as a sort of distinct object, or, for that matter, breaking down 19873 as ( 198)(73) rather than (19873) or (1)(9)(8)(7)(3) or anything practically useful.

The thing is that we can, in a sense, learn better “tokenizations”, in the sense that a 4 year old learning to read sees letters, while a 40 year old reading a novel “sees” whole words or even groups of words. The GPT architecture can’t change its tokenization scheme.

I think that discussing both sides of a cost-benefit calculation will always sound like “downplaying” one side if you happen to think that one side of the calculation is obviously and straightforwardly the weightier. The way a utilitarian would approach this (not my preferred framing but not a bad one here) would be to say “how many deaths did vaccine misinformation cause, in expectation, probabilistically?” which is very hard to estimate, and then “how many deaths would result from the elimination of Enlightenment free speech norms qua norms?” which is even more impossible to even estimate. If we have a nuclear war because anti-war speech is suppressed (again: a thing that has happened historically) and three billion people die then “obviously” speech suppression was more harmful than vaccine misinformation. But this is the kind of thing that will only be obvious when it is too late.

To put it succinctly, you think I’m minimizing the harm of vaccine misinformation, and I suspect you’re minimizing the probable harms of normalizing speech suppression, and maybe we’re both somewhat right about the other.

You’ve misunderstood and somehow taken me to be saying the inverse of what I actually wrote, so thoroughly that it is difficult to see how you could be making a good faith criticism, though it is always possible that I just wrote my post very poorly. What I said was that the deaths caused by speech suppression are abstract while the deaths caused by speech acts can be very clear and obvious. Because the harms of speech suppression are hard to see, it is harder to argue against speech suppression. In this case the speech acts related to vaccine denialism can be clearly argued to have resulted in deaths, while it is harder to argue for the danger of deaths and other harms caused indirectly by speech suppression.

Though now that I think on it, there was suppression of anti-Iraq-war speech in the early 2000s. Maybe if those perspectives had been more widely heard, that war wouldn’t have happened.

In any case, what you do in your last paragraph is merely an example of the thing I’m talking about. Misunderstanding someone’s argument and calling that misunderstanding a strawman is itself a strawman. 5.7 million deaths on one side of the ledger, okay. Well, the establishment was wrong about masking early in the pandemic, so how many more lives would have been lost if pro-masking voices had been suppressed then?

This is a very common question, and understandably so. It can be surprisingly hard to defend a general principle on ethical grounds. Principles are justified by observed regularities. Good principles do not necessarily lead to correct actions in 100% of circumstances.

The principle of “don’t interfere with the free exchange of ideas” is one such example where it is particularly hard to see the justification. The consequences of speech suppression are usually subtle —- how do you quantify the impact of a conversation that didn't happen? And at the same time, the consequences of allowing hatespeech or misinformation are usually not subtle, but rather immediate and relatively obvious.

I think we’re actually lucky because the Joe Rohan example does contain examples of good justifications. For the past two years, Rogan was openly discussing politically inconvenient inconsistencies, contradictions, and in some cases retroactively apparent lies by the establishment. It is good to know when your leaders are incompetent, when they are lying, or even when they have made a mistake and never taken responsibility for it! It is good that these conversations happened and good that people know this information.

But even if we can see that this is good, it’s still one of those abstract good-for-society things that is hard to quantify and hard to weigh against the consequences of its counterfactual non-existence. It is thus almost impossible to balance against the harms of, for example, supporting vaccine skepticism. But that’s not an argument for suppression, merely an acknowledgement that this is hard to think about clearly.

But it is a fallacy to say: Joe Rogan’s vaccine skepticism is getting people killed, therefor he should be silenced. The reason is that living in the sort of society that would silence Joe Rogan also gets people killed. The deaths are just abstract deaths.

Put it this way: image we get into a war in five years because Rogan and all the other independent voices like him have been successfully soft-pressured into not talking about the downsides, while the mainstream propaganda mouthpieces have been selling the war hard. Thus, we create a concrete terrible outcome because we allowed a culture of speech suppression. Unfortunately, even if that does happen, it will not be obvious that speech suppression led to war, because, again, general principles tend to have more subtle but far-reaching consequences.

There are also well-known edge cases where speech suppression is probably obviously the right thing to do. I personally think that the genome for smallpox should never have been published, because try as I might, I can’t imagine a scenario where the harms of suppressing the specific knowledge outweighs the possible (even likely) consequence of its dispersement. You can have a principle while being cognizant that tradeoffs can be valid.

Again, though, the reason why free expression is weirdly difficult to defend, is that the harms of “bad” free expression are usually immediate and clear, while the harms of speech suppression are abstract and distant.

Yeah, I don’t think anybody except thr OP has ever complained about Dune not feeling visually weighty and powerful, so it almost comes off as a disingenuous criticism of the film. OP is comparing stills from two movies and essentially pointing out that one is obviously digitally color-graded and the other is obviously filmed on celluloid and not color-graded. The difference is obvious in a side-by-side comparison. However, if you actually watch Dune beginning to end, your brain accepts the color palette as part of the visual language, you don’t even notice it.

I prefer the word “rival” because you can even be good friends with a rival, while the connotations of “nemesis” don’t quite work for that.

From the title my first thought was that this would be about hiding a low-power radio broadcasting station on the far side of the moon to help people reboot civilization after a dark age.

It is simply a mistake to use constructions such as “utility gain for the gifted son from living in the suburbs would be larger than the utility gain for the disabled son from living in the city”. Utility is first-person relative to the decision-maker. Your assessment of the utility of each outcome is dependent on your subjective preferences relative to the options. You, the decision-maker, do not, and cannot, know the subjective utility values that your two sons would assess to their own outcomes.

Subjective utilities are not fungible, you cannot ask Son A how utility he expects and then compare that number with the answer Son B gives you.

Thus, if you prefer A to B, yet you somehow wrote down that the utility of B is higher than A, you just made a mistake somewhere. Utility is just a scalar valued encoding of subjective personal preference. If you are using it in some other way, e.g. pretending that you are accurately measuring the subjective utility of your sons (rather than your own subjective preferences) than you are going to get weird and usually useless answers.

Some people do use utility in such a way that they assign numbers to other people’s well-being, but doing this always leads to unresolveable paradoxes such as the Repugnant Conclusion, because it’s just not how people think and decide, nor should it be.

I just figured that T1000 would never say “get out of the way!” instead of just shooting him.

You could write the same article in reverse: how many neurons does it take to simulated an ANN? A lot, because neither system is a good basis for representing the other. You need a lot of tanh functions to fit a spiking neuron and a lot of spiking neurons to represent a single tanh. This fact doesn’t say anything about the “computational power” of either system.

I am all starting to feel like it is borderline malpractice to talk about neuron/silicon performance comparisons without acknowledging that perhaps the driving design constraint in wetware is energy usage, and considering the implications this has for optimal architecture.

Sure. He structured it the way TV shows are structured, in ways that novels are not usually structured. But that doesn't mean he was imagining it would be made into TV.

>> The show was absolutely amazing when it could follow the books. The definitive show everyone talked about. The moment the books ran out it started to go downhill. And when they couldn't easily extrapolate from the books the show fell apart completely. The showrunners and writes for the show weren't good enough to make Game of Thrones on their own. No offence, but that takes a lot of skill and luck. They were good enough to adapt it though.

George R.R. Martin is still underrated, even given his success. He wrote for television for a long time and structured his novels more like seasons of TV, with each chapter serving as an episode. This is part of the reason for their popularity. Martin is also a once-in-a-generation genius of character, structure, and plot - a self-evident truth that is rarely spoken aloud because genre fiction is "not literature."

The creators of the GOT show really never even realized that they were adapting something carefully, painstakingly honed by a genius craftsman. They really figured they could just keep throwing in dramatic, violent twists and we wouldn't know the difference. In that light, it's shocking that they even made the first seasons as competently as they did.

When I worked at a national laboratory, it was always evident that research priorities lagged funding priorities. No scientific program director would allocate significant resources into an area that wasn't either already funded or practically guaranteed to be funded. I don't how NASA and JPL operate organizationally, how their budgetary decisions are made, but I would guess that no movement toward re-orienting around Starship (or any new commercial technology) will happen until funding agencies dictate that it happen.

It seems to me that responsible way to think about this sort of calculation would be to include something like QALYs lost due to premature death caused by fossil fuels on one side of the ledger, and QALYs gained due to cheap energy (supporting the development of advanced medical technology and food abundance) on the other.

It has always been a problem that the mental trick of weighing costs and benefits doesn't come naturally to human beings, but I've definitely been noticing it more and more during this era of COVID, wherein the idea of tradeoffs and cost/benefit, which should be central to any pandemic response, is rarely invoked.

I had heard that one of the major drawbacks of tokamaks was the incredible temperatures lead to situations where the smallest mechanical failure will lead to an explosion of hot, radioactive gas. Is that not the case?

I am curious whether you have ever tried applying EMDR therapy, or something related, to your bad memories. I don’t have a photographic memory but I have a reasonably good one, and I found putting in the effort to just sort of … strip away the negative valence of bad memories to be a worthwhile project.

It is unintuitive, but equatorial locations are usually stable and relatively moderate. They will tend to be uncomfortably warm year round, but never experience the temperature swings of locations much further north or south.

https://images.app.goo.gl/P4YtYRLvT5XVcn168

Look at this map of max temperatures and note that the equator runs through a band of relatively low max temperatures: Indonesia, northern South America, and the more southern part of Africa show bands of conspicuous coolness. The hottest max temperatures are in places like southern North America, India, Northern Africa, and Australia; locations that are in fact very far away from the equator.

The max temperatures of a place is caused by many things, such as proximity to water and local wind currents, but one driving reason is relatively simple: because of the tilt of the earth, every day on the equator is 12 hours. North or south of the equator, on the summer solstice, you can get 15 hour days, depending on latitude. The summer is a period of continuously intense solar heating. Again, there are other factors, but the sun is the driving force.

The thing that you notice almost immediately is that the vast majority of people suck at the medium of Clubhouse. It’s like scrolling through an endless list of the worst podcasts in the world.

Facebook is hardly DOA.

I have recently seen a sincerely incredible improvement in my migraines since I started taking supplements such as B vitamins, Magnesium and CoQ10 aimed specifically at facilitating mitochondrial function and other adjacent metabolic processes.

Something that is often overlooked in these conversations is the elementary schools should perhaps be considered completely differently from high schools. There were zero transmissions between students at my kids elementary school this whole year, despite the school opening as quickly as possible and despite several kids with asymptomatic COVID showing up at school and only being detected belatedly. I don’t think the same outcome would necessarily be expected at a high school.

This is relevant because younger kids need more supervision and are less likely to spread COVID, while older kids need less supervision and are more likely to spread; therefor keeping older kids home and sending younger ones in might be totally rational. But it doesn’t seem like this gets brought up.

If I suspect there are invisible people who can make money off of pictures of me taking a dump and can phase through doors, then I indeed might not close the bathroom doors. These metaphors never work because encrypting your text messages is qualitatively different from quotidian intuitions about privacy.