quiet_NaN
18d ago
I could tolerate LLM drivel if it was just one paragraph articulating the argument a user wants to make. Here, it is six lengthy paragraphs. LLMs as insightful as Scott Alexander, so I am not going to read a wall of text by them.
quiet_NaN
20d ago
Windows for work, and Linux for pleasure.
Funny, for me it is the opposite. I use Linux for work, (plus for some pleasure like posting here, some light gaming) and Win for pleasure (gaming only).
Second, I saw a lot of posts about how Linux eventually corrupts NTFS drives if you try to use them for anything other than simple data storage.
Personally, I would not use NTFS outside of windows much. Storage is cheap enough that getting a separate partition or even disk per OS seems preferable. In any case, you want some backup strategy. Some people will use a NTFS partition to share data between the operating systems (as linux support for ntfs is better than windows support for ext).
With regard to rust (which I have yet to learn), my understanding is that it is that the aim of the language is to provide an environment which can be as safe as Java with no runtime overhead over C, with a particular focus on thread safety. This means that you have to bother a lot annotating the lifetime of a reference in places where C/C++ would just be 'whatever, enjoy your nasal demons if you get it wrong and do a use-after-free'.
Of course, not every piece of code which is safe can be proven to a compiler to be safe. Rice's theorem and all that. Or it might be that you need to call some C function or do I/O using a device mapped to memory. Thus unsafe.
I think that from the mindset of an auditor looking for security vulnerabilities, there is still a giant difference between C and rust. In C, every line is potentially dangerous, and you need to spread your attention wide. By contrast, unsafe blocks stick out. Now, it might be that the author decided to put everything in unsafe blocks or equivalently have a function to write to arbitrary memory addresses which he calls in a zillion places, but then you can just report that that code is generally unsafe. If there are just a few unsafe blocks, you can think long and hard about when these blocks are entered, why they might be correct and so on.
quiet_NaN
20d ago
I think that the US political system is bad (FPTP, EC, special interests and all that), but not anything near maximally bad. For generations, politicians have bent the rules to benefit their side, gerrymandering, running smear campaigns, voter suppression and all that, but while the political institutions were full of infighting, they were generally playing by certain rules. Respecting the constitution, the vote of the people (no matter how much you worked to mislead them before they cast their vote) and peaceful transition of power were all part of that.
As seen in J6, Trump is not playing by the same rules. I am not sure if he even has a concept of objective truth, but he certainly decided that as he did not like an outcome in which he lost, it was false and hence the product of fraud. He then also made that terrible bid to overturn the certification of the election using his mob, which was breaking very much with mos maiorum. Luckily, his hare-brained scheme did not work.
While the institutions certainly engaged in lawfare against him, they were also not willing to break their core principles to get rid of him. The deep state certainly did not murder him, nor did 'they' commit election fraud to defeat him. When he was elected, the Biden administration let power transition peacefully -- either because his handlers could not even think of a different way to behave or because they knew very well that the same institutions which will stop Trump from permanently grabbing power would also have stopped Biden.
quiet_NaN
22d ago
To me, the workaround generally seems to be to compare your version of the document with that of a colleague. If it is identical, then there is nothing to indicate which of you leaked it. Otherwise, one might create two lossy versions which are indistinguishable.
(In general, you could add more complex watermarks, where any k copies would have bits which are the same for all of them, but used for watermarking outside their group. That would enable you to tell "Bob, Alice and Eve conspired to scramble all the bits which were different in their respective copies." However, once the conspirators know what kind of features you are using for watermarks, they could just use a technique which adds enough noise to the document to generally smear out the watermarks everywhere.)
Not breaking NDAs (unless you feel you have a strong civic obligation to do otherwise) is probably wise.
quiet_NaN
22d ago
I do not think that the people whose motivation is to use their civil service positions to influence politics are likely the ones who would jump at the opportunity of a generous retirement package.
In general, running a first world government in a tolerable way requires many more specialists than a running a resource extraction (be it oil, fruits or cocaine). Sure, Trump can put people whose primary qualification is being MAGA believers in every political position, but that is not enough to see his policies enacted. The Dems (and the pre-Trump Reps, to some degree) can draw from a large pool of PMCs who are politically aligned with them, DEI and all that.
By contrast, I doubt that Trump could staff the federal bureaucracy with Trumpers without lowering government efficiency. Sure, there are likely enough people in ICE willing to deport migrants, and the military-industrial complex will also not give him trouble while he stays within the constitutional limits with his orders, but plenty of other agencies will try to resist.
quiet_NaN
22d ago
It’s as if these people are trying to do a Weimar Republic speed run.
While I agree that murdering Trump would be terrible for the US, I want to point out that the Weimar Republic was basically cuddling up to right-wing extremists. "Oh, he lead a Putsch attempt in which a few cops died. But look at him, he was provoked by all these bloody centrist forces. And nobody who feels as German as he does could possibly be a bad person, so let's just give him the tiniest slap on the wrist to teach him to respect the rule of the law when taking power."
quiet_NaN
22d ago
second or third job (in Reddit parlance, a J2, J3, etc..)
Now I get why that J6 crowd was so angry.
quiet_NaN
22d ago
It might be worth putting this in the rules of the CW posts.
Personally, I think that using AI on themotte is bad, mentioning it is ok (if it is short and to the point). So if a comment about an AI and its behavior in a CW context ("Deepseek claims XX, this shows that the CCP is willing ..."), that is fine with me. If it is something the poster could have researched themselves, then it should mostly be verboten (or at the very least highly rate-limited and restricted to known posters). Anyone can make a motte-bot which writes more text than the real users together, and I do not think any human would like to read that (and as you mentioned, if that is their kink, they can always ask an LLM directly.)
quiet_NaN
22d ago
To be fair, "Top graduates now see AGI research as higher-status than finance or civil service." does not seem very controversial to me. It rhymes with "more kids want to be astronauts than lawyers".
quiet_NaN
22d ago
I think you are being a little unfair here.
After all, if I write up a dream, I have to put in effort proportional to the length of the text divided by my typing speed at the very least.
quiet_NaN
22d ago
Well, I think 90% of the people who write python and can do matrix multiplication can implement arbitrary-sized matrix multiplication in python. (And if I posed a similar problem, like "combine two matrices like in matrix multiplication, but instead of summing the products, just take the maximum", they would likewise come up with a solution, not simply say "that is not in numpy, hence it can't be done".)
By contrast, I would estimate that the fraction of people who grok MM and use excel which could implement the modified problem is less than 20%.
Of course, this is mostly an argument about semantics. I would not call someone who sets an alarm clock on their mobile a programmer, even though they are changing the behavior of a computer system to suit their needs. I don't use spreadsheets, but I occasionally use the unix sed command, which is technically Turing complete. However, like 98% of sed users, what I do with it is not something I would call programming -- I use it for trivial string/RE operations, but do not know how to implement a while loop in it (and also consider it ill fit for anything which requires a while loop). So sed is technically a PL, but most sed users are not sed programmers. Likewise postscript: technically Turing-complete, but the subset of users who use it to express arbitrary computation as opposed to just having it draw glyphs and graphics primitives is almost a null set. Likewise, HTML+CSS+User clicking.
Now, you are correct that Excel has had recursable lambda expressions since 2020, which together with lazy evaluation allows arbitrary computations in a single cell. However, it is my understanding that Excel was popular in the corporate setting even before that, and I would wholeheartedly recommend VBA (*) over writing a recursive lambda function in a spreadsheet cell (with very few (tiny) exceptions).
I will concede though that this is arguing over a definition. If one defines programming rather loosely, then most programs might be invocation of a Photoshop filter. Demand a bit more expressiveness, and Excel is the top PL. Demand still more (like "most users have created a non-halting program either on purpose or by accident at least once"), and Python might be most popular. Put in a bunch of conditions by machine code fetishists ("it only counts if it is compiled to opcodes and run on a physical CPU"), and likely C wins. M-x butterflies and all that.
(*) Now that are five words I had not thought I would utter in that order.
quiet_NaN
29d ago
You are correct in that -- the voting public would be willing to tolerate foreign civilian collateral damage in a way they won't when the victims are US citizens.
However, foreign operations face different challenges. I gather it is harder to get good intelligence on the operations of the Taliban than on the operations of some US based drug lord. Arrests of low tier enemies give can give you insights that you can't get from just bombing people.
The common theme on the domestic 'war on drugs' and a potential war on the cartels is that the number of people who are willing to risk their freedom and lives to make a fortune selling drugs when there is an opening in the market seems near endless.
quiet_NaN
29d ago
I think that the incentives of the cartels are pretty unlike the incentives of jihadist Palestinians.
In particular, provoking the opposed government into a violent response is good strategy for jihadists, but a terrible strategy for the cartels.
When the cartels start firing rockets over the US border, you have my blessings to bomb their launch sites (if Mexico is unable or unwilling to put a stop to that).
When forces of the Mexico government overcome your border defenses and start massacring your citizens, you have my blessings to turn Mexico into Gaza in a futile attempt to enforce a regime change.
Neither provocation will happen. The cartels like the status quo, especially the part where they are not getting bombed. The US government (presumably) likes the part where they are not criticized for accepting tremendous civilian casualties in their military operations (for a change). This is one of the cases where military confrontation would make things worse for both sides.
I mean, if you want to task the CIA with taking out cartel bosses without killing innocent bystanders (think snipers, not hellfire missiles), I guess I would be ok with that? I am however not positive that it would work, seems like even the hellfire variant did not really defeat the Taliban.
I think that less flashy responses, like trying to limit the amount of arms smuggling from the US into Mexico, plus having law enforcement counter the cartels when they operate on US soil, are likely more effective to counter spreading cartel influence.
Excel is rarely called this but it's probably the most popular programming language ever made
In a very technical sense, you could indeed argue that Excel can be viewed as programming language. It seems to be Turing-complete: I can describe the rules defining state transitions in the first cell, use the rest of the row to store the state of the band and then tell it to fill out subsequent rows, running the TM row by row. It would be up to the user to stop adding more rows once it has reached a halting state, though. Of course, the memory overhead would scale with the number of total steps.
While I am sure that there are power users (or nerds who like a new challenge after brainfuck) who implement prime factorization, array sorting or iterative solvers in Excel, my estimate is that most users only create programs which take input which can be reasonably be considered one- or two-dimensional and runs for a time which is proportional to the input size. Like, they can calculate the sum, average or minimum of a column. Ask them for anything which is above linear runtime, like the median or matrix multiplication (without resorting to purpose-built functions), which even a novice programmer who has the concept of cascading for-loops can solve, and they will be stuck. (Of course, there is always VBA, but if VBA makes Excel a programming language, then it also makes Word a programming language.)
Despite this, spreadsheets have their uses (if you can stomach interleaving the code and the data). For example, I used one to catalogue my discoveries when playing Book of Hours without spoilers.
TL;DR: calling Excel "the most popular programming language ever made" is like saying that pictograms are "the most widely read script on the planet". Sure, only a 1.5 billion people can read the word "airplane", while likely more than half of the world population would recognize the ✈ symbol. However, these universal pictograms do not qualify as a script because they are not expressive enough. Seeing people use Excel for tasks which would clearly call for a programming language is like seeing a six-year-old who insists the he does not need to learn his letters because he can just chat with unicode symbols.
Edit: also, to the degree that spreadsheets empower users, letting Microsoft take credit for that seems a bit like giving Toyota the credit for providing mobility to millions of Americans. The empowering thing is the underlying invention, the fact that the customer selected Excel and Toyota instead of SuperCalc and Fiat is of much less concern.
quiet_NaN
29d ago
Ok, sorry for misunderstanding you, it seems we are in agreement.
Seems like having two quotation lines will result in them being stuffed in the same paragraph, so I fixed my comment to make it seem less like I am misquoting you.
Rockefeller viewed Standard Oil as his real contribution to humanity. [...]
Bill Gates folllowed that model.
As some of you might know, Gates made his fortune leading a company called "Microsoft" in the 80s and 90s.
It is somewhat of an understatement to say that the Microsoft of that era was not generally seen as a force for good. They were the most hated company on the net in that era. Their software (i.e. Windows 95, 98, ME) was garbage. Its security was atrocious. Their marketplace behavior was anti-competitive. Their 'innovations' were things like a fucking talking paperclip which would try to distract you in word.
Basically, I stopped hating them when I stopped using their crap for serious work. Recent iterations of windows are tolerable as an operating system for a gaming-only machine, though.
I think that Gates has reached net good with his philanthropy, but to claim that MS was his real contribution to humanity is absurd.
Edit: made misinterpreted quotation more accurate.
quiet_NaN
30d ago
This is a classic trick. You see, either one of two cases must be the case:
- There is a longest sequence of zeros in pi with length N. In that case, L={w in {0}^n | n<=N}.
- Any sequence of N zeros will eventually appear in pi. In that case, L={w in {0}^n}.
In either case, L is decidable by a rather trivial Turing machine. That we do not know which one is the case is not an obstacle, a genius might prove either one tomorrow and then any kindergardener could built the TM.
An actually undecidable problem would be L = {x is a Turing machine | x terminates}. Here, we can't neatly split the world into two cases and just profess ignorance about which world we live in.
(Original thought: it does not feel like the outcome could be independent of whatever axiomatic system you use -- pi's digits being very computable no matter if you like ZF+C or plain ZF or Peano or whatever. (Provable theorems about the digits might be more elusive, though.) However, I would not claim that this is obvious. After all, BusyBeaver(8000) eludes ZF, when my naive assumption would have been that surely, the termination of a TM should be the same under any axiomatic system which can define it.)
quiet_NaN
30d ago
As a baseline, the us military budget in FY 2023 was 820G$. So projects of this size are neither trivial nor something which will take generations to pay the debts for.
However, I think your comparison is apples to oranges.
In the one case, we had a system of companies which had not had adequate risk management, instead trying to become 'too big to fail', at which point the US gov would have to bail them out or risk economic collapse.
In the other case, we have the US gov spending (wisely or foolishly, I am open to arguments) in the hope of gaining future benefits.
In short, it is the difference between giving your kid 100k$ for opioids to prevent them from going into withdrawal versus giving them 100k$ to spend towards getting an engineering degree. The amount of money might be similar, but the incentives it sets and the likely long term outcome are very different.
quiet_NaN
1mo ago
I think that there are solid reasons why democracies have developed cultural antibodies against Nazi aesthetics.
Now, I get not leaving everything which has ever been used by Nazis barren. For example, there are only so many two-letter acronyms, not using SS for something useful seems a bit of a waste, so I am fine with you Americans abbreviating social security thus. Likewise, just because some Neonazis like 18 and 88, even the Germans will not go all tetraphobic (yes, I know) on these numbers.
On the other hand, if someone were to say that we should reclaim "Heil Hitler", because after all, what made the Nazis bad was not their greetings but their genocides, I would reply that this is a terrible idea. It would be like saying that glowing cigarette butts -- unlike wildfires -- are generally not dangerous to humans, and thus we can throw them away without a care.
Of course, Musk's salute is a lot more ambiguous than "Sieg Heil", but in general I think that the gesture should better stay in the cordon sanitaire for a few more centuries.
My theory is that Musk did this deliberately to troll people. I disapprove.
(Also, I re-read SSCs parable of talents from 2015 recently. Scott mentions Musk no less than nine times, as an example of someone who is clearly gifted and doing good on a scale most of us could never hope to do. With a decade of hindsight, this reads somewhat bizarre -- sure, Musk did some great stuff, like making electric cars cool and establishing reuseable first stages, but he also did quite a few things which do not seem worth emulating.)
quiet_NaN
1mo ago
I don't get how the use-mention distinction is relevant here?
From what I can tell, Musk is giving some speech related to the inauguration, not a lecture on the symbolism of fascism or something. Not that a professor doing a Nazi salute for the purpose of demonstration would be very appropriate either, typically you would show a video of the period.
Even if you plead parody use, that is still use, not mention.
quiet_NaN
1mo ago
Well, it certainly looks deliberate to me. The flat hand to the breast seems a bit strange to me, I would have imagined that you make a fist first?
quiet_NaN
1mo ago
If the exploited has sufficiently unreasonable expectations, and the exploiter does or should know about them, yes, it's exploitation.
I think that is the crux of our disagreement.
In my model of the world, the woman on tinder likely has a realistic estimate of how rare it is that white men marry their tinder dates. After all, she is likely in contact with other women who are applying the same strategy, and knows how many Westerners they had sex with without getting married by any of them. She likely has some mid-status life and job in her home country (it is hard to invite Westerners over if you are living in a street or in a room with ten family members, after all). She enjoys being part of the hookup culture, and preferring white dudes is simply optimizing for the unlikely case that a hookup nets her a long term boyfriend (whom she would prefer to have a Western passport).
From what I can tell, in your model, the woman on tinder is desperately looking for a ticket to the west, in the same way that someone who sinks all their disposable income and then some into lottery tickets it trying to win the lottery. Like that gambler, she is totally deluded about her chances. She despises having meaningless sex, but carries on regardless, always convinced that the next date will finally be the one, and always being heartbroken when the guy leaves in the morning.
I think that we can both agree that having sex with someone one knows to be in the latter situation so one can save the costs for a hostel would be exploitative. I also maintain that having sex with the former woman is not exploitative.
The reason I consider the latter situation somewhat unlikely is that it basically is contrary to how women traditionally try to attract high quality mates, which is making a credible effort of appearing to be hard to get. If you are 25, on tinder and willing to fuck a man you have just met, that man can likely make an educated guess at the number of partners you had before him. While I am sure that there are men who tend to fall for woman who had tons of partners, I would assume that the average man would be slightly less likely to consider a long term relationship given that information. For example, getting hired by a Westerner as a tour guide for some token amount, being a bit flirty but not having sex with him in the first week, while also spending a lot of time with your mark seems a lot more likely to net you a boyfriend than just fucking your way through tinder. But what do I know.
quiet_NaN
1mo ago
I had the understanding that tinder is used by people looking for sex. But perhaps I am ignorant.
I would argue that it is indeed rare that the motives of people are 100% aligned. If person A hires person B as an uber driver, the shared baseline expectation is that B will transport A in a safe manner and A will pay the pre-agreed fee. If you ask A "what would be a 99th percentile outcome?", they might reply that to meet the trip would have to be quicker than expected, and B would delight them with good conversation. If you ask B, they might say if A gives them a 50% tip. While the 99th percentile outcomes might coincide for both participants of the transaction, it likely won't.
Or take a man who buys a woman a drink in a bar. Both of them have a prior probability estimate that this will not end with a "thanks for the drink" ten minutes later. In most cases, the estimates of the nonstandard outcomes (sex, marriage, becoming the next Bonnie & Clyde, whatever) of both participants will not coincide. However, this does not make their deal unfair. Even if the woman knows beforehand that the outcome the man is hoping for is not in the cards, she is under no obligation to give him a warning that she is not in the mood for sex / would not fuck him if he was the last man on earth / has vowed never to marry again / is strictly against gun violence. This does not make her an exploiter. The line I would draw is intentional deception.
Again, this is a matter of social conventions, which are of course somewhat arbitrary. I could imagine some weird culture where the buying of a drink is equivalent to marriage vows being exchanged, et cetera. Now, if the white backpacker is in a country where 80% of tinder dates lead to marriage, and knowingly flouts this convention by planning to go on tens of tinder dates without marrying anyone, then I would say that he is taking advantage of his partners.
quiet_NaN
1mo ago
So your position is that if two people have sex, but their idea of what a 99th percentile good outcome might be (say, "he falls in love with me and marries me, so that I can move to the West" vs "she brings another hot girl along and we have a threesome"), exploitation is taking place? By that test, every human interaction is exploitive. (How do we determine who is the one being exploited? Easy, whoever is higher on the woke totem pole.)
I mean, if we twisted that scenario a lot, saying that the guy is happily married in an open relationship, but falsely indicates a willingness to marry some woman in the third world so that he gets to fuck her, then sure, that would be exploitive.
I find your comment deeply unethical, but I won't substantiate why. Instead, I want you to either admit that you are a nihilist who does not believe unethical behavior is a useful category or otherwise lay out in detail a coherent theory of ethics and argue why your comment is in fact ethical. See what I did there?
I do not think that this was the intend. I mean, it would be trivial to put black bars over the naughty bits and then you can share the photo on prime time TV.
Personally, I think that different settings have different expectations of privacy. A bedroom is different from a private party, which in turn is different from the streets, which is then different from big entertainment spectacles. If you run naked across the field in the middle of a big soccer game, you can hardly argue that your privacy has been violated when people take pictures.
More options
Context Copy link