Looks like we made it. Yes, the non-AI world still exists. Bad Governor New York Governor Kathy Hochul has gone rogue and betrayed New York City, also humanity, declaring a halt to congestion pricing a month before it was to go into effect. Her explanation was that she spoke to workers at three Manhattan diners who were worried people would be unable to drive to them from New Jersey. Which,
> What are the load-bearing posts of our time? Only one way to find out. Recommended thread if you haven’t yet. I am sad you can’t easily find all the quote tweets
For games, I highly recommend Slipways if you haven’t tried it yet. Games don’t take long (30-40min), there’s a very high ceiling on getting things to fit together, you have to make decisions under uncertainty and figure out how to make the best of the web your previous actions have created. The UI is also very well done, I found it viscerally satisfying.
Fair point - definition from Google is "treat (a sacred place or thing) with violent disrespect." which I guess kind of matches (might quibble on the "violent" part). I feel like using the word Desecrate is kind of the Worst Argument in the World for this. As a Brit, I don't hugely think about Stonehenge and don't consider it particularly sacred.
If I piss on a tombstone, I desecrate it, and people could rightly be outraged, even if I cause no damage to it.
I'm not saying I buy the argument from desecration, but as a matter of political strategy the objection seems to be that desecrating what many people deeply hold sacred is counterproductive (I take no stance on this).
"as a matter of political strategy the objection seems to be that desecrating what many people deeply hold sacred is counterproductive "
I agree 100%.
I found it slightly jarring to see Zvi take the position that "this desecration is bad in itself" (which is only MY read of his IMPLIED position), that's all.
Of course it is bad in and of itself, it is not their's to do anything with, leaving aside the complete disconnection to any tangible goal they might have had other than sheer publicity.
Yeah, it's an annoying disruption but "desecration" is highly questionable in this case. I suspect a single hailstorm does a lot more damage to Stonehedge than a bit of cornstarch paint.
I am hoping people are finding these useful. I personally am happy they exist as I find long articles much more accessible through audio, and I hope at least some people are similarly inclined.
Your worst take is not to say that some advocates want to ban or replace conventional meat. Of course they do. The bad take is to predict that anyone will successfully ban or replace your meat with a subpar product in the foreseeable future and, on this basis, to blame the advocates instead of the people doing the ban. You can say you're not in favor of a ban, which I guess kudos for that, but that does not make your making light of what advocates are fighting for any less worth critiquing.
"on theory that the threat of Le Pen and the far right winning means he will win."
This is the assumption under which Macron has worked since 2017 and his only strategy to secure a win then and in 2022. The success of the far-right is exactly what you would expect was going to happen as a result of Macron spending seven years fanning the flames of the far-right to secure his position. He has no plausible deniability. This is all on him.
Take the AI analogy where it needs to go knowing this. This doesn't make Macron the savior; it does not make the AI analogs our saviors either.
A bit of gossip I heard: in addition to the years-long delays, now that the Hayes Valley TJ's is finally open, it's...not even particularly busy? Granted, summer is always the slow season for grocers, don't count anyone out until the holidays, but...the location seems suboptimal. So easy to miss the sole turnoff and then whoops, you're on the highway. Funnily enough, the second delay with the developer shenanigans meant having a lot of unused capital fixtures lying around, which is why a bunch of existing SF stores all mysteriously got infrastructure upgrades last year (generally things aren't replaced until they absolutely have to be). Never let a crisis go to waste.
I do find it funny that the same general group of people who'd propose banning "chain grocery stores" are also likely to lament the perils of "food deserts", which I'm sure academics will pinpoint the location of any day now.
My single favorite cocoa-price fact: Hershey for decades advertised a 5-cent bar; as a result of that price stickiness, the Hershey factory had dozens of different molds for the bar varying _only by thickness_ so that the bar would always turn a profit, regardless of whatever the price of cocoa was...(https://content.time.com/time/subscriber/article/0,33009,939919,00.html)
The standard libertarian position is that any regulation restricting freedom of employment contracts is prima facie suspect, since if employees really want [sick leave/safe working environment/no non-compete/etc.] they can just negotiate that into their contract, taking a reduction in their headline salary to compensate their employer for the change.
The economist Robert Frank has a good argument for why that’s generally suboptimal: Headline salary is hugely important for positional status comparisons in a way that isn’t true for anything else. (“A wealthy man is one making $100 a year more than his wife’s sister’s husband.”) This positional dynamic unlocks collective action problems in every single thing that can potentially be traded away for a higher headline salary.
If you let them, individual workers negotiating freely will voluntarily trade away other parts of their compensation package to get a bigger headline salary, in a way that is individually rational but not welfare maximizing in aggregate. This would include not having a non-compete.
I also used to use Likes to find cool new things on Twitter but also the mechanism I used to do this was Tweetdeck which X killed ages ago, so I've been living without it for a while now.
My objection to noncompetes being fair game via the "a contract is a contract is a contract" reasoning is that there's not usually parity in negotiating the terms of employment contracts. If I wanted to strike the noncompete clause and send the contract back to a potential employer they have all the leverage they need to spend less than a few seconds looking at it before saying kthxbye. I could maybe be convinced, but I haven't heard a really good argument for leverage disparities in contract negotiations being fine vis a vis "it's a contract and you're agreeing to it"; this very frequently isn't a real choice.
Steam sale reportedly starts later in this week and I perhaps-unwisely picked now to finally get around to quasi-re-playing P4G (think I played the original c. 2008) but definitely wondering if I should've grabbed SMTV +/- the new one, have heard a lot of good things about it.
The problem with ads is that, if you would pay more to get rid of ads, then you're a more valuable customer to show ads. They can make more money by charging you to get rid of the ads and then somehow showing you ads anyway. (Not enough people ever cancel anything to make this a losing strategy.) There is no equilibrium here.
Doordash was keeping drivers' tips for quite a while before the outcry was loud enough to make them stop. It wouldn't surprise me if Uber randomly kept some fraction of drivers' tips just because they can get away with it. (Comcast does stuff like this for sure.) On the other hand, part of Uber's original value proposition was that you didn't have to tip. COVID caused the memetic evolution of the idea that tips should substitute for the social safety net, which is not quite the worst possible outcome, but it's close. Related treadmill: people on hacker news talking about how they prefer Waymo because Ubers are dirty.
But at least in theoretical fields like the ones where I have experience (quantum algorithms, etc), the minimalist grant proposal system really WOULD be a massive improvement over the current trainwreck, and I can explain exactly why. Most grant proposals end up being a farce, for the simple reason that “if you had any idea what you were going to discover, you would’ve already discovered it.” So people are forced to waste vast amounts of time crafting narratives that sound good but are often totally untethered from reality, then judged on their skill with those narratives (and in fulfilling all of the funder’s formatting and other bureaucratic requirements). Experienced reviewers mostly ignore all of that and simply ask: has this person done brilliant work in the (recent) past? Are they at least vouched for by people who have done brilliant work? This is not a reductio ad absurdum, but probably the least bad way the system could possibly work.
viz., getting them to give you the money to do part 1 of what you want to do is subject to considerable scrutiny, but assuming you complete part 1 to their satisfaction, when trying to get the contract extension to do part 2 (and then part 3...), you have the advantage that they know part 1 was succesfully completed...)
Funding bodies probably could go further in this direction, of using relatively light-weight approval for extending long-running research programmes with a proven track-record.
My personal rule of thumb for CS research projects is to write down what you think will be needed, and then double it. "Why a factor of two?" asks one of our grad students. "Why not more than two? or less?" I am going to "constant of two is empirically established based on comparing predicted to actual time scale of previous projects." i.e. the random stuff you didn't anticiapte is roughly in proportion to the stuff you did anticipate.
It's probably one of those heavy-tailed distributions, so ocasionally you are going to lose badly.
e.g. it was really most entertaining when FreeBSD worked fine on single core CHERI, but console output died on multi-core CHERI. Yes, obviously, some sort of concurrency bug in operating system code. Project budget did not have a specific costed item for tracking down that problem that appeared out of nowhere and was really quite expensive to track down.
(The interrupt controller routes UART interrupt to one core. When interrupt fires, interrupt hander disables the interrupt and schedules a task to deal with it. Some time later, task runs, deals with the UART, and re-enables the interrupt. Question: does this task run on the same core that handled the original interrupt? What happens if the interrupt remians disabled on the core that is routed the interrupt by the PIC, but is enabled on a different core? Fixed now. I offer this as an example of the kind of unbudgeted random stuff that hits you out of nowhere)
There is a meta point that in computer science research, you are building things that have never ben built before, which puts you in a bad place for predicting how long they will take to implement.
> What are the load-bearing posts of our time? Only one way to find out. Recommended thread if you haven’t yet. I am sad you can’t easily find all the quote tweets
I may be misunderstanding, but quote tweets are here - https://x.com/Wangleberry/status/1793657650837942617/quotes
or here for the followup - https://x.com/bobsredmilf/status/1796601956481499288/quotes
Very fun scrolls.
For games, I highly recommend Slipways if you haven’t tried it yet. Games don’t take long (30-40min), there’s a very high ceiling on getting things to fit together, you have to make decisions under uncertainty and figure out how to make the best of the web your previous actions have created. The UI is also very well done, I found it viscerally satisfying.
Yep, Slipways is very good and I have it in Tier 1.
RE: Stonehenge - they used Cornstarch which washes off easily, see ref.
So the amount of "desecration" is questionable, as is "affecting an endangered species".
"Crime" - yes probably, though crime =/= immoral
ref: https://theconversation.com/stonehenge-protest-if-you-worry-about-damage-to-british-heritage-you-should-listen-to-just-stop-oil-232934#:~:text=The%20orange%20cornflour%20has%20been,lichen%20living%20on%20the%20stones.
The meaning of "to desecrate" is not "to permanently damage".
Fair point - definition from Google is "treat (a sacred place or thing) with violent disrespect." which I guess kind of matches (might quibble on the "violent" part). I feel like using the word Desecrate is kind of the Worst Argument in the World for this. As a Brit, I don't hugely think about Stonehenge and don't consider it particularly sacred.
Many people do (hence the outrage)?
If I piss on a tombstone, I desecrate it, and people could rightly be outraged, even if I cause no damage to it.
I'm not saying I buy the argument from desecration, but as a matter of political strategy the objection seems to be that desecrating what many people deeply hold sacred is counterproductive (I take no stance on this).
"as a matter of political strategy the objection seems to be that desecrating what many people deeply hold sacred is counterproductive "
I agree 100%.
I found it slightly jarring to see Zvi take the position that "this desecration is bad in itself" (which is only MY read of his IMPLIED position), that's all.
Good chat with you so far!
Oh yeah, I don't think desecration is intrinsically bad, though it is distasteful, imprudent, and in many cases disrespectful. YMMV.
Of course it is bad in and of itself, it is not their's to do anything with, leaving aside the complete disconnection to any tangible goal they might have had other than sheer publicity.
Yeah, it's an annoying disruption but "desecration" is highly questionable in this case. I suspect a single hailstorm does a lot more damage to Stonehedge than a bit of cornstarch paint.
I guess intent matters too
Podcast episode for this post.
https://askwhocastsai.substack.com/p/monthly-roundup-19-june-2024-by-zvi
I am hoping people are finding these useful. I personally am happy they exist as I find long articles much more accessible through audio, and I hope at least some people are similarly inclined.
It would be really funny if those workers sued the governor for defamation. Sure, call me a nazi but you dare say I’m against congestion pricing??
Lol'ed at Phillip Marlowe. I think Chandler actually wrote a country-house pastiche, though damned if I can remember what it was called.
(He went to Dulwich within a couple years of P. G. Wodehouse, incidentally.).
"English Summer: A Gothic Romance". So not quite right, but still pretty funny, considering.
Your worst take is not to say that some advocates want to ban or replace conventional meat. Of course they do. The bad take is to predict that anyone will successfully ban or replace your meat with a subpar product in the foreseeable future and, on this basis, to blame the advocates instead of the people doing the ban. You can say you're not in favor of a ban, which I guess kudos for that, but that does not make your making light of what advocates are fighting for any less worth critiquing.
Your take on Macron, OTOH, is spot on.
"on theory that the threat of Le Pen and the far right winning means he will win."
This is the assumption under which Macron has worked since 2017 and his only strategy to secure a win then and in 2022. The success of the far-right is exactly what you would expect was going to happen as a result of Macron spending seven years fanning the flames of the far-right to secure his position. He has no plausible deniability. This is all on him.
Take the AI analogy where it needs to go knowing this. This doesn't make Macron the savior; it does not make the AI analogs our saviors either.
> Not buying it: #4. Never retire. Maybe tell people different, sometimes?
I'm guessing the value in this is that it's a very compact way of expressing that
1) I am successful (because I have made enough money to retire)
2) I am available (I have a lot of free time, so if there's a cool project you think I'd be a good fit for, out with it!)
3) I am no longer part of the grind (don't bother me with LinkedIn networking bullshit)
A bit of gossip I heard: in addition to the years-long delays, now that the Hayes Valley TJ's is finally open, it's...not even particularly busy? Granted, summer is always the slow season for grocers, don't count anyone out until the holidays, but...the location seems suboptimal. So easy to miss the sole turnoff and then whoops, you're on the highway. Funnily enough, the second delay with the developer shenanigans meant having a lot of unused capital fixtures lying around, which is why a bunch of existing SF stores all mysteriously got infrastructure upgrades last year (generally things aren't replaced until they absolutely have to be). Never let a crisis go to waste.
I do find it funny that the same general group of people who'd propose banning "chain grocery stores" are also likely to lament the perils of "food deserts", which I'm sure academics will pinpoint the location of any day now.
These newsletters are what enable to keep Twitter uninstalled, so much appreciated.
re: divergent comment feeds, the video describing it is a TikTok, but the offending site is Instagram.
My single favorite cocoa-price fact: Hershey for decades advertised a 5-cent bar; as a result of that price stickiness, the Hershey factory had dozens of different molds for the bar varying _only by thickness_ so that the bar would always turn a profit, regardless of whatever the price of cocoa was...(https://content.time.com/time/subscriber/article/0,33009,939919,00.html)
Re: Non-competes
The standard libertarian position is that any regulation restricting freedom of employment contracts is prima facie suspect, since if employees really want [sick leave/safe working environment/no non-compete/etc.] they can just negotiate that into their contract, taking a reduction in their headline salary to compensate their employer for the change.
The economist Robert Frank has a good argument for why that’s generally suboptimal: Headline salary is hugely important for positional status comparisons in a way that isn’t true for anything else. (“A wealthy man is one making $100 a year more than his wife’s sister’s husband.”) This positional dynamic unlocks collective action problems in every single thing that can potentially be traded away for a higher headline salary.
If you let them, individual workers negotiating freely will voluntarily trade away other parts of their compensation package to get a bigger headline salary, in a way that is individually rational but not welfare maximizing in aggregate. This would include not having a non-compete.
I also used to use Likes to find cool new things on Twitter but also the mechanism I used to do this was Tweetdeck which X killed ages ago, so I've been living without it for a while now.
My objection to noncompetes being fair game via the "a contract is a contract is a contract" reasoning is that there's not usually parity in negotiating the terms of employment contracts. If I wanted to strike the noncompete clause and send the contract back to a potential employer they have all the leverage they need to spend less than a few seconds looking at it before saying kthxbye. I could maybe be convinced, but I haven't heard a really good argument for leverage disparities in contract negotiations being fine vis a vis "it's a contract and you're agreeing to it"; this very frequently isn't a real choice.
Steam sale reportedly starts later in this week and I perhaps-unwisely picked now to finally get around to quasi-re-playing P4G (think I played the original c. 2008) but definitely wondering if I should've grabbed SMTV +/- the new one, have heard a lot of good things about it.
The problem with ads is that, if you would pay more to get rid of ads, then you're a more valuable customer to show ads. They can make more money by charging you to get rid of the ads and then somehow showing you ads anyway. (Not enough people ever cancel anything to make this a losing strategy.) There is no equilibrium here.
Doordash was keeping drivers' tips for quite a while before the outcry was loud enough to make them stop. It wouldn't surprise me if Uber randomly kept some fraction of drivers' tips just because they can get away with it. (Comcast does stuff like this for sure.) On the other hand, part of Uber's original value proposition was that you didn't have to tip. COVID caused the memetic evolution of the idea that tips should substitute for the social safety net, which is not quite the worst possible outcome, but it's close. Related treadmill: people on hacker news talking about how they prefer Waymo because Ubers are dirty.
Awesome roundup as usual!
But at least in theoretical fields like the ones where I have experience (quantum algorithms, etc), the minimalist grant proposal system really WOULD be a massive improvement over the current trainwreck, and I can explain exactly why. Most grant proposals end up being a farce, for the simple reason that “if you had any idea what you were going to discover, you would’ve already discovered it.” So people are forced to waste vast amounts of time crafting narratives that sound good but are often totally untethered from reality, then judged on their skill with those narratives (and in fulfilling all of the funder’s formatting and other bureaucratic requirements). Experienced reviewers mostly ignore all of that and simply ask: has this person done brilliant work in the (recent) past? Are they at least vouched for by people who have done brilliant work? This is not a reductio ad absurdum, but probably the least bad way the system could possibly work.
DARPA is quite fond of contract extensions ...
viz., getting them to give you the money to do part 1 of what you want to do is subject to considerable scrutiny, but assuming you complete part 1 to their satisfaction, when trying to get the contract extension to do part 2 (and then part 3...), you have the advantage that they know part 1 was succesfully completed...)
Funding bodies probably could go further in this direction, of using relatively light-weight approval for extending long-running research programmes with a proven track-record.
My personal rule of thumb for CS research projects is to write down what you think will be needed, and then double it. "Why a factor of two?" asks one of our grad students. "Why not more than two? or less?" I am going to "constant of two is empirically established based on comparing predicted to actual time scale of previous projects." i.e. the random stuff you didn't anticiapte is roughly in proportion to the stuff you did anticipate.
It's probably one of those heavy-tailed distributions, so ocasionally you are going to lose badly.
e.g. it was really most entertaining when FreeBSD worked fine on single core CHERI, but console output died on multi-core CHERI. Yes, obviously, some sort of concurrency bug in operating system code. Project budget did not have a specific costed item for tracking down that problem that appeared out of nowhere and was really quite expensive to track down.
(The interrupt controller routes UART interrupt to one core. When interrupt fires, interrupt hander disables the interrupt and schedules a task to deal with it. Some time later, task runs, deals with the UART, and re-enables the interrupt. Question: does this task run on the same core that handled the original interrupt? What happens if the interrupt remians disabled on the core that is routed the interrupt by the PIC, but is enabled on a different core? Fixed now. I offer this as an example of the kind of unbudgeted random stuff that hits you out of nowhere)
There is a meta point that in computer science research, you are building things that have never ben built before, which puts you in a bad place for predicting how long they will take to implement.