That analogy would hold considerably more tightly if nuclear plants were things that we built to study radiation phenomena, often in convenient proximity to naturally occurring radiation phenomena, because we wished to study a matter of concern. We don't(since the fuel is so dense we tend not to even bother siting the plants in proximity to the ore; since access to coolant and position on the grid tends to matter more), and naturally occurring radioactivity only barely reaches the (scientifically interesting, but not of much practical relevance) threshold of self-sustaining chain reactions in that one uranium concentration in Gabon.

When the whole point of virology labs is the study of viruses of relevance to humans and organisms relevant to humans(livestock diseases, etc.) either because of direct pathogenic potential or because they have scientifically interesting properties; it starts to look way less like a crazy sinister couldn't-possibly-be-a-coincidence that, gosh, the virology lab was totally doing work on a class of viruses with some concerning and interesting behaviors known to exist in the vicinity.

Doesn't mean that they didn't fuck up; the higher biosafety level facility standards and procedures are really fiddly for a reason; but when the whole reason you'd pump research funding into something is so they can poke the viruses that make you nervous the fact that they've got a bunch of vials of mean-looking viruses onsite just seems way less automatically damning.

I think that this might be a difference in preferred phrasing: "Coordinate the population" could mean Glorious First 5 Year Plan, Comrade!; but "we're going to do rule of law with laws made this way and cases under the law handled this way; and the existence of a free market shall be secured by these restrictions on force, fraud, deception, and assorted tortious shenanigans" is also coordination of the population, as are maintaining a military and a foreign policy posture, administering a currency, or uniformity of weights and measures. Even the most optimistic 5 year planner doesn't coordinate fractally deep; and even the lightest-touch small governor tends to desire the facilitation of independent coordination rather than seeing a 'sit back and watch the anarchy' policy to be something you do as an end in itself.

It seems like it requires something a bit weirder than mere lack of values.

There are certainly lies you can tell that straightforwardly advance your agenda if you are indifferent to truth, or people you can kill or allow to die to suit your purposes; but what baffles me is why this would be a case where that works:

If your problem is that the US COVID numbers were both bad in absolute terms and strikingly bad in relative terms vs. comparable countries; how exactly does it help you to claim that a perfidious chinese bioweapon did it? Is "I, the president, manifestly bungled the response to a novel pathogen emerging from a lab in a country epidemiologists have long been jumpy about in terms of respiratory infection threat risk(mostly because of the year-in/year-out surveillance of what flu is going to get up to between humans, poultry, and pigs this time; but it's not like MERS or SARS were ancient history)" actually better than "I, the president, manifestly bungled the response to a novel pathogen of zoonotic origin that originated in a questionably sanitary meat market"?

From a purely US public health perspective it's basically the same fuckup(with not a little depraved indifference) in either case. From an exercise-of-presidential-power perspective the possibility that it is sinister chinese is arguably the worse of the two; because now there's potentially more that some combination of State Department diplomatic efforts and DoD/Intelligence Community intelligence efforts could have extracted directly from the lab that allegedly produced the virus and might have already had useful information about it ahead of time to be either cajoled into sharing or simply espionaged out by one means or another.

That's what baffles me. Trump lying as easily as he breathes? Must be one of those days ending in 'y'. Trump voluntarily going to a lot of trouble to tell his own supporters that he didn't bungle the response to a tragic natural event; he bungled the response to negligence, possibly even a dash of malice, from a not-wholly-friendly foreign power? How does that benefit him?

I don't really grasp the logic, or deviation from logic; but it seems like it has the shape of a sort of weird 'value' system of its own: one where there's some sort of mysteriously capped supply of culpability(rather than exactly as much culpability as is required by the scale of deviation from sensible process); and if it can be pinned on some other malefactor it must therefore be the case that others are innocent despite their role in the process. I just don't get it.

People who tell exculpatory lies are liars; but they are the sort of liars I can at least intellectually understand: someone accuses you of guilt, you don't want to get punished for guilt, so you claim to be innocent. Not morally upstanding; but the logic checks out. Here, though, it is being treated as though it is an exculpatory lie; but there appears to be no exculpation implied. It's just baffling.

I suspect that, on average, it is a thing that skews mac without strictly being linked to it.

Apple tends to distribute its build quality and production values a bit more evenly across its models(bad things have definitely happened; but it's not like there's a designated 'cheap plastic shit' line that exists to offer 'starting at just $399.99!' at all costs; and when the bad things do happen they are often as not a design or component problem that takes out one of the expensive ones; rather than it being that macbooks get the hinge design that saved a nickel and maybe lasts 2 years while macbook pros get the hinge that had actual mechanical engineers rather than glue); but that(along with their relatively early move into soldered RAM and storage; and their ongoing battle of thinness vs. being realistic about CPU TDPs) means that a lot of the surviving models in good shape with otherwise reasonably attractive attributes (decent build quality, pretty solid panel choice for the time) are unexpandable base model versions. Hooray.

It's not like the equivalent equip-your-kid-for-college-for-less PCs of the era would be any more compelling today; but the really lousy ones are a lot more likely to either be dead or have had corners cut on something like panel resolution to a degree that makes them hard to justify caring about; while the real top shelf hardware of the period is substantially more likely to survive and be usefully upgradeable because that's part of what made it the good stuff back when it was made. Whatever random Inspiron was on special the day you bought the W510 would...probably not...be worth going to much trouble to get back into operation.

The part that does seem to be more mac-specific, or at least more relatively visible, is people doing stuff that suggests they actually liked the OS that the hardware shipped with, or one close to it. There's some of that for classic UNIXes as well, though on a smaller scale. On the PC side actually liking windows 98 is mostly not a thing(retrogame enthusiasts sort of do it; but are typically focused pretty specifically on the application they care about and how to give it what it wants; not so much on actually liking its original habitat); and, unlike the constant tinkering to try to retain macOS compatibility when Apple drops it; most of the NT-focused tinkering is on how to bludgeon the shell back toward 2000; while the people who are just running linux instead exist entirely orthogonal to the original OSes trajectory.

So 'uncensorable' and announced at a 'blockchain' event; do we think that they'll go the route of actually being on chain, to technically deliver something adjacent to the promise at enormous cost in actual usability; or will that be too much effort and it'll be a standard whatever-cloudflare-will-put-up-with effort with minimal interest in moderation but no actual technical resistance to 'censorship' as soon as it's someone posting kiddie porn in ways that actually put the operators on the hook, rather than just internet assholes being what they are?

Did I miss the memo where the alleged 'efficiency' guys are also handling 'fairness'? Or the wave of lunacy that would make someone want to just sit in the dark without CVE information until the UN Ad-Hoc Working Group for The Cyber does something; because we wouldn't want to risk having someone who isn't paying potentially benefiting from our defense program?

I'm not sure if this is from the 'penny wise pound foolish' side or the 'barbarians are now inside the gates and setting things on fire' side of the current administration; but it's shockingly stupid to just scrap an initiative aimed at curbing the (significant) costs of security problems for the US, private and public sector, for want of a way to paywall the Europeans out of it or whatever.

It seems a bit premature to interpret this as an actual signal of either Google's progress or their belief in the state of the possible given how much enthusiasm there is for AI hype; and how cheap, as a marketing exercise, this hire is going to be.

It is possible that it's sincere; but it would also be a trivially obvious hype move: for the cost of some sort of 'futurist' on payroll you can imply, without specifically lying, that your AI progress is so scary good that you totally need a post-AGI theorist to handle it.

This is the sort of thing that would get written up as a clever(if probably not at the top of the urgency list) inferential attack if you could do it against remote hosts or VMs under the same hypervisor. Apple has had the temerity to declare the comparatively low bitrate of the leakage to be a privacy feature. Love that Cupertino attitude.

What's great, of course, is that they are doing all this in a black box, whose security vs. people who aren't them they guard fairly jealously; and they are making no particular claims about exactly how fine-grained their inferential attack will be; or whether they will ever feel like turning that dial in the future.

We already know that they will turn the dial if they feel the need, since they previously didn't do this and now their vaunted 'synthetic data' is not good enough and so they do; but the technique offers an almost arbitrary range of possible resolutions, from exceptionally low granularity(but unlikely to produce much of interest) to "it's not technically exfiltration if I guess 1 or 0 and the remote host tells me if I'm right or wrong; rather than just copying the data directly"(which is basically copying the data directly; only less efficient).

I assume that they will make some effort to avoid the truly trivial case; but this vague and fancy description will be...flexible...cover for an extremely wide range of possible levels of intrusion; and they will no doubt smarmily insist that they are on the side of low granularity; but have no reason to necessarily actually be there(especially when the behavior needn't be uniform across devices; all sorts of options for sampling or targeting present themselves).

I can't believe that senile belligerence only makes everything feel like a position of strength. Crazy stuff.

I suspect that Nvidia isn't too concerned: reports are that the TSMC facility in the US is relatively close in cost to the ones in Taiwan; and one area of computer 'manufacturing' that has been most likely to retain some US presence has consistently been relatively late stage integration of expensive and configurable datacenter gear: not super sexy, actual option cards and CPUs and RAM and such get manufactured elsewhere and a warehouse in Texas just has people screwing in the ones you ordered so the vendor doesn't have to guess at demand for every configuration variant enough in advance to ship them by boat and retain acceptable lead times.

Plus; if Nvidia is offering a high-visibility win; they are much, much, more likely to be allowed to do a little fudging around the edges for anything that would actually be expensive or highly disruptive(eg. the TSMC facilities in the US do diffusion but not advanced packaging and testing; I'm guessing that nobody will be hassling Nvidia about that; or about any Taiwanese parts they may need to rush in to meet a particular deadline).

Doesn't necessarily mean that the cost is zero; but, especially when you are talking a very high margin product and a very politically sensitive business dealing with a very volatile president with an expansive view of his own powers(both in terms of tariffs and in terms of Nvidia's ongoing fights about what parts they can sell to who and whose fault it is that China never seems too fussed about getting cutting edge Nvidia cards); this seems like a very cheap concession that is likely to buy they a number of...useful...chips that can be cashed in to handle other matters that are likely to arise.

My impression is that universities aren't so much the sole class of institution charged with the responsibility; but the main one that quite overtly embraces the objective and does what it can to provide clear signaling mechanisms like academic degrees and clear 'do you want to learn about X? Take X 202: A Not Complete Layman's Examination of X" course offerings.

It's not like anyone believes that this is the only way to learn things; or nobody would ask for 'years of experience' or document or look at employment history on resumes; but private employers absolutely(with some limited exceptions in the very academically modelled corners of some R&D operations, often the ones cushioned from immediate organizational pressure by some relatively juicy profit center or a lot of cultural clout, Bell Labs style) do not have the interest that universities do in providing externally useful signaling information, or in providing learning opportunities that don't align with their fairly immediate requirements.

Depending on the university and the company; the company's actual assessments might actually be more honest and accurate(since, while the noble theory demands that teachers evaluate students; the constant whispers that students are customers who are paying for degrees and must be passed can corrode the cultural norm that educators must uphold standards; while, however often they fuck it up, companies theoretically have a very strong self-interest in knowing how competent their employees are; and at what); but there's no equivalent incentive for them to share that information. They can't avoid indirectly leaking some data; if Company Z was willing to cut me a check every year to do Computer Stuff for them; that's implicitly a statement about my abilities in Computer Stuff; but it's not like I can just tell Company Y "Sure, check out my performance reviews in Computer Stuff at Company Z"; while the University of City Name will, absolutely, be happy to verify that I got a Degree in Computer Stuff with intermediate honors and here's the transcript.

Universities absolutely don't have anything resembling a monopoly on being places where people learn things; but they are highly atypical in their role (sometimes realized; sometimes a bit aspirational; but at least clearly understood) of being independent of the student, only a shit degree mill will outright talk you up just for cash; but aligned with the student rather than with itself when it comes to providing readily understood credential signaling to 3rd parties. It's not like all employers love onerous non-competes and wage suppression cartels; but they aren't just banging on the door to formalize a standard for externally recognizing their employees for work experience and on-the-job learning so that others can more efficiently poach them.

I suspect that billionaires(or even millionaires, or even not-3rd-world-slum-dwellers) don't like to think about survivorship bias; but when the question is how other people should be trained I doubt they find the odds all that concerning.

Thiel would presumably prefer to get someone actually punchy who feels like he owes him out of this(consider, for example, the utility he gained by setting up JD Vance as a relatively minor, but rather more successful than he otherwise would have been, VC bro) and at least a couple of reasonably photogenic cases who sell some bullshit consumer facing startup and can be feted on techcrunch as worthy rebuttals to filthy liberal education; but people who are incorrect about being startup successes have their own uses; specifically being employees or contractors for people who are correct on that point. Normally you don't pay for their education, so he presumably wants better than average for these ones; but it's not like they need to just be scrapped if they don't hit the Forbes 30-under-30.

The question of whether a high school teaches 'statistics' or "statistical analysis" sounds a lot like one that hinges on how optimistic the person writing the CV or the course catalog is feeling.

If you didn't have a stats course(not necessarily a function of rigor: there are definitely the schools were low-theorem geometry and precalc get dragged out more or less indefinitely to provide a supply of low threat math; but also ones where the implicit goal seems to be beelining calc and linear algebra; where geometry and statistics get sidelined in order to have more time to go in that direction) you definitely aren't doing 'statistical analysis' (no, that one time in science class where you got carefully coached through declaring a result statistically significant probably doesn't count); but describing your stats background as 'statistical analysis' is making it sound a lot more ready for 'analyst' purposes than it likely is by the end of senior year.

Presumably if the point is to find eager optimists who are not wholly ignorant of the foundations of the subject that's fine; but if there is anyone calling you an 'analyst' for high school stats it's probably one of the shops where they hand out titles rather than raises.

Sssh. The fact that 'meritocracy' is an aristocracy that makes the occasional economic diversity hire, rather than the radical utopia of social mobility, is supposed to be the quiet part. We called it 'meritocracy', so obviously now there are no barriers except lack of merit; which means anyone who opposes us must be jealous losers.

Obviously American Greatness doesn't actually have any ingredients aside from belligerent insistence; so none of that is a problem.