By pulling ten million people a year from farms into factories and ploughing 40% of GDP into infrastructure and education. Sounds like a sound analogy to me.
I agree, and cloud compute is poised to become even more commoditized in the coming years (gazillion new data centers + AI plateauing + efficiency gains, the writing is on the wall). There’s no way this makes sense for most companies.
The advantage of renting vs. owning is that you can always get the latest gen, and that brings you newer capabilities (i.e. fp8, fp4, etc) and cheaper prices for current_gen-1. But betting on something plateauing when all the signs point towards the exact opposite is not one of the bets i'd make.
Well, the capabilities have already plateaued as far as I can tell :-/
Over the next few yeas we can probably wring out some performance improvements, maybe some efficiency improvements.
A lot of the current AI users right now are businesses trying to on-sell AI (code reviewers/code generators, recipe apps, assistant apps, etc), and there's way too many of them in the supply/demand ratio, so you can expect maybe 90% of these companies to disappear in the next few years, taking the demand for capacity with them.
> Sadly, instead of having better laws we get fallacy rhetoric by people who probably have never touched, much less fired a gun in their lives.
Why is this the litmus test for being qualified to write gun legislation? Do we also expect our lawmakers to have tried heroin or downloaded child porn so that they can regulate those activities?
This is a bad example. I've been notionally pro-ownership but also pro-regulation my whole life, and one of the major problems with gun legislation in the US is that it's incredibly poorly written and does not reflect the technical reality of guns.
The government allows private ownership of automatic weapons, but hasn't issued any new tax stamps for 50 years. You can convert any semiauto gun into a full-auto gun for a few cents of 3D printed parts (or a rubber band). The hysteria over "assault weapons" basically outlawed guns that _looked_ scary, while not meaningfully making anyone safer.
I think yes, it is reasonable for Congresspeople to fire a gun before they legislate on it, because otherwise they are incapable of writing good laws.
Good gun regulation in the US would probably look like car insurance, where gun owners need to register and insure their weapons against the possibility of crimes being committed with them. There are so many guns compared to the amount of gun crime that it would probably not end up terribly expensive, especially if you own a gun safe.
> The hysteria over "assault weapons" basically outlawed guns that _looked_ scary, while not meaningfully making anyone safer.
This wasn't the goal by the congresspeople, and that them having fired a gun would've changed that goal.
That was the goal. They knew they weren't going to be able to pass any kind of legislation that actually msde people safer, but they wanted to look like they were "doing something".
This is incredibly common. It's the primary reason behind the TSA and its continuous expansion, for example.
> It's the primary reason behind the TSA and its continuous expansion, for example.
I'd also add that the TSA is a good reason why we shouldn't expect talking legislators to gun ranges would make better gun laws.
The reason the TSA is what it is is because legislators fly more than most people. If you've ever been to DC you see a lot of this sort of security theater everywhere.
So much of the TSAs budget should be redirected towards what would actually make long distance travel safer, improving the ATC and Amtrak.
Thats defacto gun registration- and worse: registration with a private entity not beholden to due process. Given current realities, anybody who registers their firearm in such a manner can expect a no-knock raid because they were nearby when somebody phoned in an engine backfire as a gunshot.
So make it allowed that the insurance is tied to the gun. You buy a lifetime policy for that serial number, provide payment, and you're done. Payment can be provided anonymously at a window in cash, if that's your thing.
If you want discounts because you live in a low-crime area, have a gun safe, have many guns, etc. then obviously the storage location for the weapon needs to be declared to the insurance company.
That's not true. They have millions of digitized 4473s. They are banned by law from creating a searchable registry of gun owners but they digitize paperwork on a daily basis.
Thanks for the clarification. I knew there was limitation placed on them to hamstring their operations under the auspices of preserving the 2nd amendment.
You're welcome to come up with a better litmus test, but it's beyond clear that lawmakers writing gun control regulation have less than a wikipedia level understanding of the topic. See "shoulder thing that goes up", the weird obsession with the Thompson, the entire concept of an Assault Weapon, etc.
Wikipedia has much better information about guns than most of the people talking about them in politics, generally speaking.
It's not too surprising, considering the way the rules are written at the ATF. There's basically zero logical thought that goes into pistol vs rifle vs felony:
ATF rulemaking can be unintuitive and arbitrary but there really is a level below it occupied by people who have dedicated a significant chunk of their lives to trying to restrict firearm ownership, who genuinely seem to believe that Die Hard, Rambo, and Spaghetti Westerns are real life. Politicians who can't answer basic questions about their legislation, who have to be told live on air that magazines can be repacked, that just make up impossible crime statistics. Yeah it's stupid that the ATF has decided that vertical grips are a rifle feature but angled grips aren't, but it gets worse.
There's no legal definition per Congress. Generally speaking, braces are intended to stabilize a pistol against your arm [0], whereas a rifle stock is meant to stabilize against your shoulder. However, braces can technically be "misused" such that the rear of the brace fits against the shoulder, meaning it is used as a stock. Likewise, the distinction is so small something as simple as a sling attachment to the stock could make it a brace, or an articulation that could be used as a cheek rest turn a brace into a stock, converting a pistol into a rifle or vice versa.
For awhile, the only way to know the difference was for the manufacturer to submit an NFA and hope.
The ATF has been in court (and lost) quite a bit [1] over this.
A "pistol brace" is designed and "intended" to be braced against your forearm to stabilize the "pistol" in a way that allows you to shoot a particularly large and heavy "pistol" with one hand. The ATF said this was fine, although I think they really regret that now.
Stock goes against the shoulder. Brace goes against the elbowpit. If you let the brace touch your shoulder, your braced pistol suddenly becomes an unregistered SBR, and you become a felon. Oopsie!
> Do we also expect our lawmakers to have tried heroin or downloaded child porn so that they can regulate those activities?
It would be nice if they delegated to experts, instead of think tanks or populism, when it came to dealing with these. Both are examples of rampant regulatory failure.
Knowing the difference between a think tank and experts might be hard without some rudimentary knowledge to spot nonsense? I don't know, actually asking. It seems to me that the primary skill we need in our leaders is that of spotting experts talking within their field and actually listen to them while ignoring others. The primary trait, which is even more important, is character so that they act on what they here in our best interests instead of their own.
Having first hand experience in contraband, and having the knowledge to create and pass laws are very different things. If you think that current lawmakers dont know enough about contraband to regulate effectively, what makes you think someone who knows loads about drug use and porn would be able to contruct decent watertight legislation?
In this specific discussion familiarity does seem relevant. I don't think shooting is so relevant, but printing and assembling are.
You don't have to be a life-long user to regulate heroin, but if you start legislating second-hand heroin smoke, people might look at you sideways. You kinda need to know a little even if you've never actually ever seen heroin. If you demonstrate severe ignorance, people are going to call you on it.
I don't think its unreasonable to ask politicians to be familiar with how the machinery they are regulating functions and is used.
To use your heroin example, this is akin to banning spoons or needles because they heard those are tools of the heroin addict. It shows a lack of understanding on the part of the regulator and has a far reaching effect on people legally using the items.
Having a clue about how guns work, or the general reality of any other field one may be attempting to legislate, is absolutely crucial. With guns it just happens that actually firing them is a good way to gain (some of) that understanding.
litmus test wise, regulators of 3d printing should be able to create strong parts with a variety of 3d printing mechanisms.
they should at least be able to understand that a 3d printer is akin to a turing machine and what the real limits are - strength of the printed material vs length of the strip of memory.
It’s more like people who barely use computers regulating software features and development.. oh wait
I don’t own a gun, and think guns should be regulated more and better, but the heroin let alone another one are just flawed. There are no legitimate, non-life-ruining use cases for either of those analogies.
The ISS consumes roughly 90kW. That’s about *one* modern AI/ML server rack. To do that they need 1000 m^2 of radiator panels (EACTS). So that’s the math: every rack needs another square kilometer of stuff put into orbit. Doesn’t make sense to me.
And what happens every time a rack (or node) fails? Does someone go out and try to fix it? Do we just "deorbit" it? How many tons per second of crap would we be burning in the upper atmosphere now? What are the consequences of that?
How do the racks (or nodes) talk to eachother? Radios? Lasers?
What about the Kessler Syndrome?
Not a rocket scientist but 100% agree this sounds like a dead end.
Communication is a well-understood problem, and SpaceX already has Starlink. They might need pretty high bandwidth, but that's not necessarily much of a problem in space. Latency could be a problem, except that AI training isn't the sort of problem where you care about latency.
I'd be curious where exactly they plan to put these datacenters... In low Earth orbit they would eventually reenter, which makes them a pollution source and you'd have no solar power half the time.
Parking them at the Earth-Sun L1 point would be better for solar power, but it would be more expensive to get stuff there.
Seasons mess that up unless you're burning fuel to make minor plane changes every day. Otherwise you have an equinox where your plane faces the sun (equivalent to an equatorial orbit) and a solstice where your plane is parallel to the sun (the ideal case).
Huh, I didn't know that that was possible without burning fuel. Kind of wild that it only works because the Earth has an equatorial bulge and isn't an exact sphere.
Suboptimal - likely. There is some utility: a green letter is more useful than a yellow. Checking for a in two locations when a is a very commonly used letter is __useful__. Still likely much more useful to check for the presence of a fifth letter than a chance at knowing more precisely the location of an a.
There seems to be a progression of Wordle strategies.
Playing with a set start word (or words, e.g. "SIREN OCTAL DUMPY" or people who go the "AUDIO ADIEU" route).
(Many people also go down the rabbit hole of looking for "optimal" starting words or choices based on the original word lists.)
Then, once you've played that for a while, you find it's not that much of a challenge unless you end up in one of the forms of madness like _A_E_, and you'll switch to playing in "hard code" (e.g. correct/green must be played again in the same place in all subsequent tries, yellow letters must also be reused each time).
The hard mode starting with the same word gets a bit boring, so people move on to varying the start word each day, either pulling them from a list or just using the answer for the day before.
There's no "correct" approach obviously, people can play the game however they want and extract the fun/anger however they want.
Because a wordle-in-one is meaningless. It doesn't mean you're any good at Wordle, the way a hole-in-one suggests you're good at golf. It definitely doesn't mean that you're a "Genius" as the game puts it, because you were operating with zero information and didn't employ any skill or intuition. It just means you burned some luck points on something that doesn't matter.
I used to use “stare” or “stale” as the starting guess when I played Wordle, thinking you’d want to start off with the most common letters, like R-S-T-L-N-E from Wheel of Fortune.
It’s this. There are many five letter words that are not “wordley”. Words such as, idk, bokeh, are technically part of the lexicon but would never appear as a solution. The wordle bot will even tell you this if you guess them — “good guess, but unlikely to appear as a solution”. The crossword has a similar sort of unwritten rule, maybe not as strict, but really hard technical words seldom appear.
I didn’t know it was called crosswordese! I wonder what the most common term used is. As a very occasional player, for some reason ARIA, IBIS, and VENI/VIDI/VICI stick out, but I’m sure it’s actually one with an E.
VENI/VIDI/VICI are easy for anyone who studied Latin (as indeed used to be common), and ARIA is similarly easy for anyone who knows about opera. Basically, the crossword is for snobs.
I agree that crosswords often include cultural references that lean towards certain demographics / assuming particular education, and that can feel exclusionary if you don’t share that background - and there's even an argument to suggest snobbery might be behind those choices.
But I disagree that that makes it for snobs. Snobbery is more about an attitude of looking down on others or their tastes, whereas knowing Latin or being a fan of opera is really just about exposure.
Sure, there exist some (too many) opera fans who would say something like "it's real art compared to pop or hip hop being low class trash", but that's not a defining part of liking opera and plenty of people who like opera aren't snobs. Ironically it's a different form of snobbery (sometimes called reverse snobbery though personally I hate that term), to dismiss anyone who learned Latin or who likes opera as being a snob!
The middle 4 are all fairly common words. "Ode" isn't super common, but I hear it in "An ode to..." phrases. And "err" I've only ever heard in 1 phrase: "To err is human."
That's not really the concept. People know what an orca is.
But if you see a crossword clue that says "black and white animal", you know that the answer is ORCA without even needing to look at the number of letters in the answer. (Could it be "skunk"? Could it be "panda"? No, those are stupid questions.) Same thing if the clue is "marine predator". (Could that be "shark"? No.) The words I listed are incredibly likely to appear in crossword puzzles. That's what's weird about them.
An épée is one of three types of sword used in the three styles of Western fencing. As such, it's about as technical as, say, the words "touchdown" or "mitt".
It's also just the regular French word that means "sword". But although crossword puzzles frequently ask you to know common French words, I've never seen one clue the answer EPEE that way.
If you took fencing at an Ivy League school for you PR requirement you would know all about foil, saber, and epee fencing. Not everyone gets to row crew.
Scrabble is a competitive game, not a puzzle, and therefore subject to a different set of constraints. (Players in a competitive game are trying to win; a puzzle author, if they're any good at their job, is ultimately trying to lose.)
In particular, you have to consider the equilibrium. If you only allow a subset of words in Scrabble, this replaces the competitive advantage from knowing lots of words that no one uses in real life, with a competitive advantage from knowing the exact contours of the border between acceptable and unacceptable words. I would argue that this is even worse; at least if you learn lots of Scrabble words you're learning something about the real world.
By contrast, Wordle can self-impose whatever constraints they want on solutions, and people don't have to know what those constraints are in order to solve the puzzle. (It can help a little on the margin, which in a perfect world would not be the case, but it's much less of a problem for the puzzle-solving experience than the Scrabble equivalent would be.)
Ya that's a good point for competitive scrabble. However today I think a lot of people's main exposure to Scrabble comes from WordsWithFriends (and recently, the new NYT games version). In those games, there's no penalty for getting a wrong word, it just won't let you play it. In that context, I at least think it would be nice to have a setting with a more limited list... it could be like Chess timed variants.
It's obviously an impossible challenge to draw those contours in language. Wordle did pretty well though! And going the other direction, just allowing everything that could possibly a word, just starts getting ridiculous.
Even in casual Scrabble-like games, I expect using a restricted set of words would create a lot of "come on, that's totally a real word, why can't I use it" moments. Most people know at least a few uncommon words that most other people don't (because it's different words for each person).
The Wordle list of legal guesses is not substantially curated; AFAIK basically all five-letter words legal in Scrabble are on it (except on offensiveness grounds, which was a highly controversial decision). If this were not the case, I predict you'd get user dissatisfaction as per above. Wordle's list of possible answers is much more curated, but that's my point; it can err on the side of conservatism, because users won't notice if a word that they'd expect to be on there is missing, whereas they will notice if such a word is not allowed as a guess.
Wouldn’t that make Scrabble only harder and more annoying to play? With that limitation you’ll get situations where you play a perfectly valid word, but it gets rejected because it’s not in the list of approved words. To get good at that version of the game, you’ll have to study the Scrabble word list instead of the dictionary.
With Wordle the limitation is only put on the words the game generates as answers. You can use obscure words to guess, they just won’t be the answer.
this is already the case with scrabble; there is a strictly defined scrabble word list that determines whether a word is acceptable or not, and it often leaves out words that you might find in some other dictionary that is not the official scrabble one (collins for most of the world, or a custom dictionary for american scrabble)
I am not a native speaker but how does your wife name the caulk in the shower? Silicone? Or do you maintain it in such pristine condition that no word was ever spoken about it?
printf() debugging is still considered a best practice in the eyes of many. I still remember being really surprised when I heard my famous (Turing award-winning) CS professor tell the class this for the first time.
The thing about printf debugging is that it works universally. All languages, all platforms, all stacks. Even down to the lowest levels of most software, there will always be some sort of log available.
While some tools/frameworks might have more robust debugging tools, if you have a dynamic role within an organization, you may not find it worth the effort to set them up if your target platform is constantly changing.
One real world example of this from my own work in PHP - there is a tool/lang-extension called XDebug that is great and provides step through debugging, but historically, it has been a pain to configure. It's gotten better, but when I can just as easily add a few `dump()` statements that expose the same data, it's overkill. Very rarely do I need to actually pause the running request to debug something and 99% of the time, I just want to know the state of that object within an specific `if()` block and a debug log gets me that same information.
reply