native_samples

In particular, most people recognize the possibility (at least theoretically) that someone can be both an extremist and also correct. Whereas by your definition that's impossible.

Governments like the idea that they can fix things via market interventions but it never quite seems to work out how they intended.

The problem is that this is the religious position, not the other one. If you go investigate this properly you'll find there's tons of scientific dispute about even very basic things like what the temperature actually was 5 years ago in specific places, let alone more complex topics like what the climate would do in response to an immediate doubling of CO2 levels (so called ECS).

And the impact of global warming itself isn't at all obvious. All the effects so far are so small you can't directly detect them with your own senses, you have to use sensitive monitoring equipment and/or accept a lot of conjecture about chance and probability. In turn that requires you to have absolute confidence in the people doing the measurements and processing the data, but plenty of people don't have that confidence for various legitimate reasons (e.g. the Climategate emails). So it's pretty much by definition not obvious.

"Believing nothing should be done about climate change does not make one an extremist"

Doing nothing is the default position in any debate so arguing for it by definition makes one not an extremist. You can't be a "nothing extremist", that's a contradiction.

Extremism takes the form of positing extreme scenarios like the end of the world, and demanding extreme solutions like massive changes to everyone's ways of life. The people who say "let's not do that" are the ones pushing back on extremism.

This is at the root of a lot of the fighting over climate change: it creates enormous cognitive dissonance in people. Those who decide they want to fight it want to see themselves as rational devotees of science, but find themselves being the guy in the sandwich board yelling "The End Is Nigh" and demanding extreme policies. And just like sandwich board guy they struggle to explain the details of how the world is meant to end. Very few people who have tied their identity to climate extremism can actually handle a debate about the details which is why they instantly retreat into screaming that it's all "settled science" (which is a lie).

The word "may" is doing a lot of work here. Lots of things may cause massive disruptions to human society. One thing we can say with certainty: the mitigation efforts being proposed to reduce CO2 levels will cause massive disruptions to human society.

That's nonsense. Temperature isn't a direct function of CO2 in the atmosphere. Even climatologists don't believe it's that simple. That's why there's so much debate about the true value of ECS, that's why they spent much of the 2010s trying to explain the pause in global warming by reference to heat absorption of the oceans, etc (they since changed the mind and now argue that they were simply failing to correctly measure temperatures for decades, which is even worse).

The false belief that climate = CO2 levels is one of the things that creates skepticism about climatology, because anyone who bothers to check the actual facts instead of relying on "high school level chemistry" will immediately see that it's false. Moreover if it were actually that simple then there'd be no need to worry because if you simply map CO2 level increases to temperature increases then you don't get anything even close to a doomsday scenario. All the "we need to reach net zero yesterday" type claims are predicated on hypothetical and quite complicated feedback loops that involve far more than just a straight mapping of temperature=CO2 levels.

As for "virtually never", why would you think "never with exceptions" is the definition? "Virtually never" here just means "it happens very rarely relative to the total number of COVID deaths". Isn't that normal use of English? We're not communicating in SAT here.

I agree that the statistics are blurry, but they're mostly blurred in the direction of exaggerating the risks. I'll take your word for it that your acquaintance did indeed die of COVID, but it must be remembered that such classifications were being handed very liberally, especially in the USA where COVID death classifications were being directly incentivized with money.

People chased the vaccines because of the promise of effectiveness. That was reasonable. Going further and making up claims with no basis wasn't.

Actually, possibly not :( This breaks down into two separate questions:

1. Are they effective at reducing risk of death / serious illness now.

2. Were they so in the past?

To answer these we can't rely on any scientific gold standards despite the importance of the questions, because the trials showed no impact on death or actually the Pfizer trial had more deaths in the vaccine group than the control group (due to extra heart attacks). This was written off as "not statistically significant" (this isn't the right way to use statistical significance as a concept). Then they vaccinated the control group, thus ending the RCT phase of data collection.

The problem was that despite the constant messaging, COVID is virtually never deadly outside of the very elderly or sick, and those people tend not to join drug trials. Thus despite having ~65,000-ish participants there were not enough COVID deaths to detect any impact on this metric. Hospitalization meanwhile and "severe illness" were likewise not trial endpoints. So only noisy observational data is available.

For (1) the data from countries that actually release the raw data shows very mixed results indeed, with deaths being higher amongst the vaccinated cohort for some dosing ranges and age groups, lower for others, and overall there is no clear message from the data that vaccines are reducing deaths.

https://bartram.substack.com/p/update-on-the-failure-of-the-...

"It appears that things have got somewhat worse since last week’s report, with the hospitalisation and death rates in the double vaccinated (not boosted) exceeding that seen in the unvaccinated for all aged over 70, and with a higher death rate in those aged over 60 as well. What’s more, there now appears to be practically no benefit of vaccination with two doses of vaccine for those aged under 60, in terms of the remaining protection against hospitalisation and death."

That's for the UK but similar numbers can be found elsewhere in the rare cases where governments actually release the actual data (normally, they hide the actual numbers). Also, what we actually care about is overall mortality because vaccines can and do kill people, and just changing the cause of death isn't very useful. There too the news is disappointing, with small increases in excess death being correlated with booster rollouts but not the winter Omicron wave:

https://bartram.substack.com/p/increased-deaths-in-england-f...

All this is official UK government data so enough reason to at minimum reserve judgement. Data in the past is unfortunately heavily corrupted by mis-allocation of people to the different status buckets. The methodology they use would calculate a big but transient impact on death and sickness from an injection of water.

Here are even more popular-but-unpopular opinions (facts).

We did not "all" think vaccines would have a meaningful impact on transmission. Sadly they never had any impact on transmission. A whole bunch of studies have measured the SAR split by vaccine status and there's no difference, that's not something specific to Omicron. Although health authorities claimed they would reduce transmission, they had no scientific basis on which to do so because the trials didn't test transmissibility, or even impact on severe illness or death. They only tested reduction in PCR positivity rates (and only symptomatic, iirc).

People who were already checking the claims of public health authorities against the trial data knew this from day one, which is one reason so many people opted out of the vaccination process: if the vaccines are so great, why are they making unsupported claims with 100% confidence? Why lie?

Sadly the answer seemed to be not for medical reasons, but because public health had become obsessed with total vaccination regardless of need. The means had become the end. Telling people the vaccines reduced transmission manipulated huge numbers of people into taking them, even if they weren't in any risk groups and knew it, just out of a sense of obligation to others. And then telling people vaccines reduced transmission enabled them to implement vaccine mandates by claiming the unvaccinated were dangerous to the vaccinated.

https://fantasticanachronism.com/2020/09/11/whats-wrong-with...

"The original IAT paper is worth revisiting. You only really need to read page 1475. The construct validity evidence is laughable. The whole thing is based on N=26 and they find no significant correlation between the IAT and explicit measures of racism. But that's OK, Greenwald says, because the IAT is meant to find secret racists ("reveal explicitly disavowed prejudice")! The question of why a null correlation between implicit and explicit racial attitudes is to be expected is left as an exercise to the reader. The correlation between two racial IATs (male and female names) is .46 and they conveniently forget to mention the comically low test-retest reliability. That's all you need for 13k citations and a consulting industry selling implicit bias to the government for millions of dollars."

https://psycnet.apa.org/record/2013-20587-001

"IATs were poor predictors of every criterion category other than brain activity, and the IATs performed no better than simple explicit measures"

One might think a site dedicated to intellectual curiousity would be interested in stories like that, or that at least flagging such stories would be penalized, but no.

When bytecode is being interpreted, every bytecode instruction has a safepoint after it. In compiled code polls are scattered thoughout the code, for example on method entries and loop back edges (so the comment above about needing to add code in tight loops to help the GC is wrong).

Then summing the 18-65 buckets (working age).

A lot of programs written in Rust would almost certainly be better off being written in Kotlin. The one in the article is a good example. Why are they writing a messenger bot in Rust? That doesn't seem like the sort of use case Rust was targeted at. It's also a modern language with a lightweight syntax and pretty good static typing, great refactoring tools etc, but it's way more user friendly. There's no borrow checker because it's GCd, compile times are much faster (at least for the JVM version) etc.

You can also use Kotlin/Native or GraalVM native-image to produce standalone executables that don't need a full JVM, if that's a requirement for your use cases. The native images are astounding. They can start faster than programs written in C and their memory usage is also way less than a typical JVM app. Downside is of course compilation time but you can avoid that by just developing on the normal JVM and then AOT compiling at the end when it's time to release.

a. The official number is ~254,000 (in the USA) which is certainly a major over-estimate. True number is certainly far less.

b. Employees who died or retired don't count as resigned.

It's really not.

You seem to be talking about COVID science in general, but this specific WHO debate is something else. Regardless of which way you slice things, it's colossal incompetence and/or mendacity:

1. Either the changes being proposed for Germany and Sweden in the WHO report are correct, in which case the WHO published a report with extremely basic mistakes in it and then realized "immediately" that they had got it all wrong. These wouldn't be "science is hard" type errors, they're "we have no idea what we're doing" type errors, like using the wrong figures or not extrapolating a curve correctly.

2. Or (seems equally plausible), the original report is reasonable and the criticisms are politically motivated. The article admits that they immediately "re-checked" their work because "some critics used the report to criticize Germany's pandemic policy". Then they decided that their scaling factors were not only wrong but shouldn't have been used at all yet they don't explain this.

The Nature article doesn't seem to give the people who made the original decisions much of a chance to explain themselves. We're left to assume it was Wakefield who did this but he also doesn't explain the sudden change of heart over scaling nor where those numbers came from or why they were used. Nature just presents what they did as objectively wrong, but it's all just modelling. So Wakefield decides that he doesn't like the fact that lockdown sceptics were using his report and rewrites its conclusions. Why is the WHO report adjusting the death figures reported by national governments to begin with? That would be interesting to know, but he doesn't explain and nor does Nature. He just discards the scaling factors and then asserts his new work is correct now (why should we believe this?).

Almost certainly Wakefield and friends don't like the original methodology because it makes Sweden look good and Germany bad. It's predictable "scientists" would attack anything that reports these outcomes because they've been doing that for two years. Sweden proves beyond doubt that lockdowns were a terrible disastrous error by our expert classes, their great death figures make the whole thing immediately obvious. Bad modelling is the primary tool public health people use to control politicians and populations, and they will happily play with models until the data says Sweden was a bloodbath and Germany saved the world.

"it doesn't alter science of excess mortality one bit to me, that it happened."

It should! You have a group of people who went through multiple rounds of review, then adjusted their methodology immediately after publication for what appears to be political reasons. Clearly this stuff is more like opinions than facts.

Which is often the right thing to do unless you have very specific knowledge of that experience e.g. a referral, because so many people flat out lie about their experience and accomplishments. If you can't directly verify it, the accuracy is so low it's best ignored.

As for your experience, I'm glad to hear you're now happily self employed (so am I!). How are you proving your skills to customers, if you can't handle job interviews? Or are you making a product?

"Have you considered that you might not have a full understanding and appreciation for human psychology and the wide variance among humans?"

I have an average such understanding, having lived an average life. But the problem remains: someone who claims to be able to do a job fine but can't demonstrate that capability when asked, is from an employer's perspective best avoided. That isn't a flaw in the interviewing process, that's the point of it! The reasons why someone can't demonstrate it are irrelevant because there's no better alternative available, and anyway, someone who can't handle the pressure of a job interview is very likely to crack in other situations. You claim it's not true and there's something psychologically unique about interviewing, but, I don't really believe this and clearly neither do other people doing hiring.

You seem to be suggesting something totally unfalsifiable - that candidates who do a bad job aren't actually bad programmers, but are just nervous inside in undetectable ways. There's no way I or anyone else can argue against this notion because there's no way to prove a negative like that, but suffice it to say that the sort of issues I'd associate with nervousness - minor slips that they then notice and correct, stumbles, brief moments of forgetfulness etc - are not why people tend to fail interviews. They fail interviews because their coding skills suck.

I've done hundreds of interviews as the interviewer. I've had people seriously panic maybe once or twice in that entire time. I've heard occasionally that the rates for other interviewers were higher (but still low), if that's the case then they just sucked at interviewing.

Where did therapists come into it? I've never heard of anyone going to a therapist to pass coding interviews, that would be ridiculous. Learning to control your emotions is something people are meant to learn as children. It's a standard skill you're supposed to have to interact with others, like basic politeness. If you're having breakdowns as adults then it's expected that the situation is really intense, like the death of a loved one, breakdown in a relationship etc. Job interviewing is a normal part of life and shouldn't be as stressful as those things.

"It's already bizarre that candidates with many years of experience in the field have to study intensely for job interviews."

The whole "cramming leetcode" thing is way overblown, or possibly a modern feedback loop in which crammers are making it harder to detect genuine skill and experience. Again I've done a lot of interviews and nobody ever mentioned having to prepare for them. That idea seems to be a relatively recent idea (last ~10 years or so) and is probably a result of so many people competing for very highly paid jobs at a small number of firms. Normal software jobs should ask people to demonstrate their skills, but it shouldn't be something that requires exam-like prep for any working programmer.

The link you posted seems to reinforce my point though. It's a reimplementation of DALL-E, not anything done by Google. That seems to be how it goes, everyone tries to duplicate the OpenAI work even though Google's models are (or claim to be) more advanced. The mindshare value of making their demos available appears to be phenomenal.

Is it reproducible - well, only for flexible definitions of reproducible. If you try to follow the method in the paper and don't get results as good, is the issue the method or that you didn't follow it well enough? You can't follow it all that closely because they often aren't detailed enough and the training sets are constantly changing.

The issue here (for Google) is a bit different. If they were keeping their research proprietary because they wanted to turn it into a product and sell it, that's one thing and totally understandable. It's very expensive and needs some way to financially justify the cost. Their justification is quite different though and raises questions about their whole AI initiative. What's the point of creating SOTA AI trained on the internet if you're afraid of what people will use it for? Their efforts to make ideologically acceptable AI don't seem to have worked yet, nor has OpenAI been embarrassed by abuse. So OpenAI is powering ahead here and everyone talks about their models, whilst Google's languish in obscurity. Like, Scott Alexander is talking about the limits of DALL-E 2 that Google claim they already solved, but it's irrelevant because Scott can play with DALL-E and not Imagen.

Unfortunately the impact of Imagen will be low like all the other Google models because, as keeps happening, they don't trust people enough to actually release it, not even in demo form:

"Imagen relies on text encoders trained on uncurated web-scale data, and thus inherits the social biases and limitations of large language models. As such, there is a risk that Imagen has encoded harmful stereotypes and representations, which guides our decision to not release Imagen for public use without further safeguards in place."

This seems to be emerging as a clear pattern - OpenAI trains an AI and then opens it to a group of select invitees (the usual suspects) with lots of opaque T&Cs that restrict what they're allowed to say about it [1]. For example, OpenAI forbids users from sharing any outputs that contain realistic faces even though the AI generates such pictures all the time, even when not requested [2]. This is reminiscent of how RDBMS vendors forbid users from publishing benchmarks and usually results in people giving OpenAI some stick for not living up to their name.

But OpenAI is still light years more open than Google, which routinely announces just months later that they've trained an AI far better than anything OpenAI produced, but doesn't provide any evidence beyond their own cherry picked examples. The justification given is always the same: the AI has learned un-woke things like the fact that certain jobs tend to be done by specific genders, and would generate examples based on that learnings. And because the Google researchers are crazy they think that seeing pictures of white male builders or young female nurses would be actually dangerous to the public, so they just don't provide any API access at all, not even to little select lists of friends.

IMO this is turning into a serious problem for AI research. It's already notoriously non-reproducible, but OpenAI's work is at least auditable. People can do experiments on DALL-E and see that it's real for themselves, explore the limits and come up with ways to handle them, as is happening here. They could in future build apps that use them. In contrast Google, who should by all rights be at the forefront of this space, keeps getting eclipsed by OpenAI again and again because they're now so woke that they've retreated into a tiny little purity bubble. They claim that they'll make their AIs available when they figured out how to brainwash them to have the right views, but they were claiming this is an "open problem" for years and never seem to release anything. So all the talk is about GPT-3 and DALL-E and Google's equivalents just get forgotten.

[1] https://www.lesswrong.com/posts/uKp6tBFStnsvrot5t/what-dall-...

[2] https://www.lesswrong.com/posts/uKp6tBFStnsvrot5t/what-dall-...