Thursday, April 27, 2023

Thursday Tweets -- Elmo and Sneetches

When we first started mocking the "great news for DeSantis" meme, we were taking a fairly controversial position. Then the winds started to shift and it turned into something of a victory lap, reminding people that we were here first. Now, though, everybody is jumping on the bandwagon, which means:

1. It's not much fun anymore.

2. Pundits are starting to move from prematurely declaring DeSantis the inevitable nominee to prematurely declaring his campaign doomed, which in black swan season is just asking for it.

So no more Ron for a while. Fortunately, we have plenty of Tucker and Elon to fill the gap.

Before we get to the reaction to the defenestration, take a minute to watch this clip and remind yourself of the level of crazy TC was feeding to his viewers.


Tuck was always quick to stand up for his country. That country happened to be Russia but the principle remains the same.








There is no one so despicable on the far right that you can't find a loon on the left to defend them.




Loads of charming details.

SCOTUS News (though not technically 'News')



As noted elsewhere on Twitter, who would have thought we'd have two SCOTUS justices mired in financial scandals and neither of them is named Kavanaugh.


 

More for the bad NYT framing file.





His family must be so proud.


Joe McCarthy -> RFK -> RFK jr. -> Roger Stone


Because Tennessee went so well for them.


Excellent thread, particularly about carriage fees.

At which point I'm contractually obliged to mention...

 

 To understand the dysfunction of the current anti-abortion movement, you have to remember that, while many of the members are motivated by deeply held beliefs about the nature of human life, probably more are driven by exposure to decades of horrifying disinformation.



 

Checking in on Elon.



 


Of all the burns delivered to Elon over the check mark fiasco -- and there have been many -- this may be the best of them all.

Elon being a deepfake would explain a lot.


The latest from ProPublica.


Back on the AI beat.


You come across the most unexpected conversations on Twitter.


I didn't read the article so I can't recommend it, but I love the clip.

Wednesday, April 26, 2023

Revisiting the Red Tsunami -- in which I (sorta) defend the pundits from Andrew Gelman

(Or at least spread the blame around)

I am, for obvious reasons, reluctant to challenge Andrew on questions about polling and political science, but over the past few years, I've spent an unhealthy amount of time studying and commenting on the coverage of elections which leads me to push back on a number of points in his recent post "No, the polls did not predict “a Republican tsunami” in 2022." [For still another take, check out Kaiser's reaction.]

 

                     We'll just set the Wayback Machine to 2015...

Though it's a bit off the main topic both for his post and this one, let's start with this:

"In the 2016 primaries, the polls were right about Trump having a bi[g] lead over his opponents; it was the pundits who were wrong."

No one hates pundits more than I do, but if you go back and review the coverage, you'll see that the even-at-the-time embarrassing push to deny that Trump was the front-runner came primarily not from pundits but from data journalists. If you're looking for flawed, narrative-serving "Trump won't get the nomination" arguments in the NYT archives, you'll find far more from Nate Cohn than from David Brooks. 

If you're a real glutton for punishment, go back and read the posts we did on the subject back in 2015. One on 538, and two on Nate Cohn. Silver later went back and made a real effort to acknowledge and address his mistakes. As far as I know, Cohn (who was a much more serious offender) never did.


                    There are polls, then there are polls.

The main focus of Andrew's post is this quote from a London Review of Books article by Adam Shatz. 

 The polls, unreliable as ever (this was one thing Trump got right), told us that high inflation and anxiety about crime were going to provoke a Republican tsunami.

 To which Andrew replied:

But that’s wrong! First, the polls are generally not “unreliable,” and it’s particularly wrong to cite Trump here (more on this below). Second, no, the polls in 2022 did not “tell us” there would be “a Republican tsunami.” The pundits in 2022 were off, but the polls did just fine.

For example, here are the poll-based forecasts for the Senate and House of Representatives from the Economist magazine (they should have some copies floating around in the LRB offices, right):


 


 The actual outcomes (Republicans ended up with 49 seats in the Senate and 222 in the House) are close to the point forecasts and well within the forecast intervals.

The polls are not magic—they won’t tell you who will win every close race, and they can be off by a lot on occasion, but (a) they’re pretty good on aggregate, and (b) no, they did not predict a Republican tsunami.

 I found Shatz's piece painfully conventional and largely free of insight (but remember, I hate pundits). That out of the way, the example Andrew gives doesn't address what Shatz actually said. This is going to be a fine point but an important one. Just so there's no possibility of missing context, here are the three paragraphs where Shatz talks about the polls. (Only the first two are relevant to this conversation; I'm including the third just for the sake of completeness.)

I spent much of the night brooding on the impending birth of the United States of QAnon. A Republican sweep seemed inevitable. The president’s party usually gets clobbered in midterm elections. Despite all Biden’s achievements – high employment, investment in infrastructure, a historic climate change bill – his approval ratings were in the low forties, as if the only things Americans could remember about him were his stammering delivery and the chaos of the (otherwise popular) withdrawal from Afghanistan. The polls, unreliable as ever (this was one thing Trump got right), told us that high inflation and anxiety about crime were going to provoke a Republican tsunami. Afterwards, Republican legislators and election-denying secretaries of state (the chief election officials in US elections) would join forces to prevent the Democrats from winning in 2024 – or ever again. The country appeared to be careening towards constitutional crisis, and the other side had all the guns.

...

All of this led Biden to give one of the most sombre and unflinching speeches delivered in recent memory by an American leader. ‘Lies of conspiracy and malice,’ he said on 3 November, had produced ‘a cycle of anger, hate, vitriol, and even violence’. The survival of American democracy, he continued, was now ‘the biggest of questions ... You can’t love your country only when you win.’ Much of his rhetoric sounded old-fogeyish (‘a struggle for the very soul of America itself’), but that doesn’t mean he wasn’t right about the threat. His political calculations were also sound. According to the polls, inflation and crime were the major issues for most Americans, not abortion or democracy. Yet Biden’s insistence that ‘democracy is on the ballot for us all’ resonated with many voters who were fed up with the persistence of minority rule, the furies (and the outsized power) of the far right and the normalisation of violence by political pundits and elected representatives. Whatever Biden’s approval ratings, his warnings in this speech may have helped to set the stage for the Democrats’ strong showing; so, too, did his success in restoring a modicum of civility to American political life. 

...

Why were​ supporters of the Democrats – including seasoned election-watchers – so easily persuaded by Republican triumphalism? The polls were one reason, of course. But susceptibility to Republican hype is more directly a result of the Trump years. There is every reason to fear that Trump, or rather Trumpism, might return. It’s easy to mock MSNBC-watching liberals who rapidly resort to analogies with Germany in the 1930s (especially when analogies with episodes in American history, such as Reconstruction and the McCarthy era, are closer to hand and more illuminating). But there is little doubt that the United States has become more vulnerable to authoritarian challenges thanks to the deterioration of its democracy over the last two decades. The reasons for this decline are many but include the absence of limits on campaign finance and the overwhelming influence of corporate money, the right-wing assault on black enfranchisement and the descent of some quarters of red America into conspiratorial culture-war fanaticism.

 In both relevant paragraphs, Shatz is talking about opinion/attitude polls while Andrew counters with results derived largely (entirely?) from electoral polls. This is an obvious mismatch but it also raises some subtle but far more significant points, but before we get to that, there are a few things we need to get out of the way.


                  When you only look at good polls, the polls are good

In the final days of the election, observers such as Rosenberg (who, like his associate Tom Bonier, came out of 2022 looking really smart) were raising the alarm about a wave of junk polls designed to juice the red tsunami narrative.

Conscientious poll aggregators either gave the junk little weight or ignored it entirely, so the damage on that side was minimal, but when we make blanket statements about "the polls," we can't simply pretend these don't exist.

 

                    Elsewhere in the Economist

At the peak of the campaign, a couple of analyses from the Economist and the NYT rattled everyone's cages.

More 2020 Poll Forebodings
By Josh Marshall
|
September 12, 2022 8:52 p.m.

Over the last couple weeks we’ve seen write-ups pointing to the possibility or probability that polls in 2022 will underestimate Republican strength just as they did in 2020. Now Nate Cohn has one in The New York Times. Cohn’s analysis is particularly interesting to me since he’s been a fairly consistent skeptic of polls in recent cycles in terms of their ability to accurately gauge the strength of the Trump coalition. G. Elliott Morris had a similar write-up in The Economist last week. Philip Bump kicked the tires on Morris’s claims in the Post. While they break down the numbers, the gist is pretty straight-forward: If you take the polling error from 2016 and 2020 and plug it into our current polls you go from Dems being in a strong position for holding and even expanding their senate majority to more like a 50-50 odds of holding the chamber at all. On the House, a GOP majority and maybe a significant one is basically a given.

So what to make of all this? On the specifics these analyses are right. And that is sobering. No question about it. So how likely are we to see a repeat of 2020? I wish I knew. The one thing that stands out to me about both these analyses is that they leave out 2018. That was obviously a good year for Democrats, though more in the House than in the Senate. It was also a midterm. So there’s a decent argument that it’s a more like comparison to this year’s midterm. Morris notes that even though 2018 was a good year for Democrats it had similar polling biases to 2016. When I asked Morris why he excluded 2018, given it’s the best year recently for Democrats, he told me that the point of his experiment wasn’t to predict the 2022 outcome but rather to show a worst case scenario for Democrats. I think that’s a reasonable answer and I suspect it’s Cohn’s too. 

Morris's statements to Marshall are more cautious and qualified than his piece in the Economist (where the phrase "worst case scenario" does not appear, which in turn was more cautious and qualified than were his tweets on the subject. None of this is to suggest that there was anything inaccurate or irresponsible about Morris's analysis, but it's another reminder that the pundits in large part got the impression that we were looking at a red wave because they were reading data journalism in places like the Economist and the NYT.

 

                     Downplaying Dobbs

A number of analyses (most notably this one from Nate Cohn at the NYT) suggested that the impact of abortion would be less than most people were expecting.

Polling suggests an overturning of Roe v. Wade might not carry political consequences in states that would be likeliest to put in restrictions.
Cohn did give himself some wiggle room but on the whole it was a bad analysis as we pointed out at the time. By way of comparison, it would have made more sense to do an analysis showing how little usable data we have on the political impact of an overheated US economy (high inflation/high employment) as compared with stagflation. I don't recall anything like that getting any attention.

But these are minor issues compared with the problems with how pundits, data journalists, and in some cases political scientists talk about polls.


                    Polls do not predict. That's the problem.

Polls (at least electoral polls) look like predictions. You can often treat them as predictions, using them as the sole input for very simple models, but the polls themselves are not making any predictions; they are just tools for measuring beliefs and attitudes, and the ways we need to think about them (and about how they can go wrong) are entirely different than the ways we need to think about predictive models.

Specifically in this case, we need to think about what did and didn't go into the models that made the predictions. As far as I can tell the big inputs were electoral polls, attitude polls, presidential approval polls, the economy, and recent/historic voting patterns. These are all standard and quite reasonable choices when trying to predict elections, and every one of them was holding up a big "RED TSUNAMI" sign.

The absolute certainty that this was going to be a disastrous year for the Democrats entrenched itself early in the election and it did not originate with the pundits; they got it from sources like the Upshot, and you can't really blame them for reading their own papers.

[Obviously, I can blame them -- as mentioned before, I hate pundits -- but in this case we need to take a hard look at the "serious" people, not just the op-ed dwellers.]

It's not like no one saw that there were potential problems with the models. A number of people (including Joseph and I) argued we were so far out of the range of observed data that nothing model based could be trusted. There were also analysts looking into the data who came up with a very different take, most notably, Tom Bonier who staked out a very lonely position based on voter registration numbers, special elections turnout, and early voting data. Ironically, his stuff showed up in the NYT as guest editorials, meaning that the opinion section of the paper actually had better writing about data than what we saw in the data journalism section.


https://twitter.com/MarkPalko1/status/1650233537659031552?s=20

Tuesday, April 25, 2023

Had to get there before the snowmelt started and you have to worry about flash floods

Finally did some hiking in the Alabama Hills.

 


 That's the east side of Mount Whitney. The estimates I've heard have the melt continuing into June. Even as far south as the San Gabriels, I'm still seeing a surprising amount of snow on the mountains.

It has been an extraordinary year.

(back to very serious blogging tomorrow.)

Monday, April 24, 2023

Ponderings about ancient Rome

This is Joseph.

So I love analogies with ancient Rome and I love the works of Bret Devereaux, an academic famous for making it clear to me how shallow my understanding of ancient history really is. But I worry about this particular quote:


I think that this must apply to Augustus (also named Julius Caesar because, of course he is) as the original Gaius Julius Caesar was a dictator but did not reign in any real sense of the word. At the time I suspect Augustus would have quibbled with the word "reign" but in retrospect I think we can allow that it was probably an accurate description of his power and influence. 

But what I think Augustus was worried about was an underproduction of elites and not a general decline in population. Both population and population density increase between 14 CE and 164 CE. Rome had a higher fertility than China, which is not evidence of a serious problem. 

In the modern west, I think we have the opposite problem in elite overproduction. That sounds like a oxymoron, but if elite training takes decades and there are insufficient opportunities at the end, people sacrifice decades of income and life for poor outcomes. Look at the issues with academics in the arts. It is a big source of discontent to have people train for ~10 years to end up needing to find a new life plan or to be trapped as adjuncts. 

There are defenders of this system, although notably the defender is from the UK where PhDs are 3 years long and so the opportunity cost is a lot less. It'd also be different if PhD was a job, itself, and not extensive and rigorous training for a job. I will also point out that most people doing a PhD work very hard -- so it is extensive effort for a goal and not necessarily a time of leisure that would make the low salary seem like a good plan. Nor is there a general sense that elites have few children in the US (Elon Musk famously has ten. Donald Trump has five. Joe Biden has four. Bill Gates has three. I don't see a fertility crisis among the elite. 

So I guess I am not certain of the truth, but I do think that shallow analogies are potentially dangerous and could push us in the wrong direction. The fertility challenge for the modern world is the middle class, and I would look squarely at long training phases, expensive housing, and limited childcare if I wanted to confront that. 

Oh, and spoiler alert. Augustus was a long, long way from the fall of Rome. Augustus dies in 14 CE, The Western Roman Empire goes through a golden age for a couple of centuries and lasts until 476 CE. The final fall of the actual Roman Empire takes place in 1453 CE, not precisely an indication of imminent disaster during the time of Augustus. 




Friday, April 21, 2023

Weekend Movies -- the all time great winter Western noir (plus a bonus flick)

You can still see some snow in the mountains, but the days are getting hot and it won't be there long. The passing of the season got me thinking about one of my favorite winter films, André de Toth's sadly under-recognized Day of the Outlaw featuring Robert Ryan and Burl Ives at the top of their games, with strong support by Tina Louise (who really should be remembered for more than that TV show). Unforgettable last half hour.

It's available free with ads from Pluto.




No snow but a lot of rain in James Whale's The Old Dark House, so I don't have a seasonal excuse for talking about it here, but it was a virtually lost film for decades. It's finally back in circulation, showing up on Svengoolie (who had been trying to get it for ten or twenty years) and on streaming services like the also free with ads Tubi.

Does The Old Dark House live up to expectations? No, but it's difficult to see how it could. Not only does it face the build-up that comes with decades of being a film everyone talked about but nobody saw; it also has to stand up to Whale's other horror films of the era, Frankenstein, Bride of Frankenstein, and the Invisible Man.
 

Still, taken on its own, it stands up quite well, I'm glad I finally got to check it off my list.





Thursday, April 20, 2023

Thursday Tweets -- that hollow thump you're hearing is the sound of a narrative imploding

And the hits just keep on coming.

While there's no question Ron DeSantis is having a cruel April, his apparent collapse has less to do with recent developments and more to do with people finally acknowledging how flimsy the original narrative was, a product of motivated reasoning and wishful analytics (which a lot of us called out at the time).

Republicans were looking (as they have been since 2015) for a way to lose Trump and hold onto his supporters. The establishment press (particularly the NYT) was looking for a GOP candidate whom it could cover without embarrassment, perhaps even one they could both-sides without inviting mockery. These strange bedfellows all convinced themselves that DeSantis was the savior they were seeking. 

I can't think of any candidate who received a comparable build-up starting not just before the election but years before the campaign. The idea Trump was basically done and Ron was all but a sure thing became a foundational assumption of the standard narrative. If the pundit class had any capacity for self-awareness, they'd be having a rough April too.

None of which is to say that DeSantis won't get the nomination -- Trump is a spectacularly self-destructive old man with a terribly unhealthy lifestyle and Ron is still next in line -- but the most likely path to the nomination looks nothing like what expert opinion was telling us a few months ago.





Another bit of bad luck for DeSantis came when Disney fired the stupidest CEO in the Fortune 500. With Chapek, Ron could come off as a Trump-style bully (red meat for the base). Now Disney is slapping him around and his response is petulant and weak.




And in other GOP news...



“I was reading a sign high on the wall behind the bar: ONLY GENUINE PRE-WAR AMERICAN AND BRITISH WHISKEYS SERVED HERE I was trying to count how many lies could be found in those nine words, and had reached four, with promise of more...” ― Dashiell Hammett



 

And they call us Arkansans poor white trash.







When defending the indefensible, always start with positive framing (also useful for access journalism).



Remember when Republicans disliked Russian stooges?





 

 Musk/Tesla/Twitter



 

Elon has always been best understood as the sophomore who keeps coming by the freshmen dorm rooms because he can bum their weed and impress them with deep thoughts from the philosophy class he audited.




Nice to find someone who's read the classics.
Speaking of things they should have seen coming.


Step 1: Have a rich daddy.

Swifter than most.


When bad things happen to stupid, annoying people.

 

Back in my teaching days, my classroom was always the one with the chess boards. I feel vindicated.


In AI news...



And for the math nerds in the audience.

Wednesday, April 19, 2023

Critiquing my own side

This is Joseph.

Sometimes it can be healthy to point out places where your political party falls short. Smart self-reflection is the basis for positive change, after all. So lets talk about the senate and entitlement. Hopefully this is a retrospective, set for being published in the future, and the events discussed are over by now.

The Democrats have the president and 51 senators. One of them has been in the hospital for depression for 44 days. One of them came down with Shingles in February and does not have a return date in mid-April.  Two others are Joe Manchin and Kyrsten Sinema, meaning that they both now hold the ability to defect and give the opposition a majority, just by not voting. 

Now I think the insane push to put democrat aligned judges on the bench is playing into the Republican supreme court gambit, but I recognize that there is a lack of consensus on better ideas, Why there is this lack of consensus is fair enough -- a diverse party will struggle with organization and strategy. 

"I am not a member of any organized political party. I am a Democrat." -- Will Rogers

So execution of strategy is important. The senator with shingles is on the judiciary committee, so her absence blocks forward progress. The Republicans have blocked a measure to substitute her out. She is 89 years old and is within 18 months of leaving the senate. 

So far, so good. What scares me is the defense of this process. See here. See more from Talking Points Memo here:
Senate Dems Look Moronic 
The most generous interpretation of Senate Democrats’ ineffectual effort to get Republicans to help them sideline Dianne Feinstein so they can resume confirming Biden judges is that they needed to go through these motions and have them fail before they could prevail on Feinstein to resign her seat.

I could barely type that without screaming into the void. Because of course that’s too generous by a lot.

Senate Republicans lined up yesterday to declare they’re not going to lend Democrats a hand. Of course not. It was silly to pretend they were going to help, both because this is the modern Republican Party and because … why should they really?

So we’re back to where this all started a week or so ago. Until DiFi resigns her seat, the Judiciary Committee is hamstrung and new Biden judicial nominations will be stalled. It’s true that Senate Democrats can’t make her resign, but the song-and-dance routine of the past week doesn’t inspire confidence that they really get what’s at stake or have coalesced around an effective path forward.

But this defense is shocking: "They all deserve a chance to get better and come back to work. Dianne will get better. She will come back to work.” To quote Unforgiven: "Deserve's got nothin' to do with it."

We saw the same issue with Ruth Bader Ginsburg who stayed until the very end and switched the median vote on the supreme court from John Roberts to Bret Kavanaugh. People talked about how tragic it was that she was participating in the court from a hospital room, ignoring that it was her decision to stay in office during the preceding 8 year democratic presidency. 

The role of an elected official is to represent their people and to advocate for them. Elite offices are not a prize to be won but they are a solemn responsibility to the constituents to be an effective advocate and representative. It is the office that has respect and not at all the person in the office.

I think we should look at these issues through this lens -- how do we ensure that people are properly represented. I think it is obvious that allowing the judiciary to keep being stacked when one side is in power but not the other is a poor strategy and likely to cause issues in the long run. 

Tuesday, April 18, 2023

Progressive journalists and bloggers are unified behind the need to champion reproductive rights, protect democrat institutions, and loosen fire safety laws

I HAVE NO POSITION ON THE SINGLE STAIRWAY QUESTION.

I HAVE NO POSITION ON THE SINGLE STAIRWAY QUESTION.

I HAVE NO POSITION ON THE SINGLE STAIRWAY QUESTION.  

I HAVE NO POSITION ON THE SINGLE STAIRWAY QUESTION.

I HAVE NO POSITION ON THE SINGLE STAIRWAY QUESTION.

I HAVE NO POSITION ON THE SINGLE STAIRWAY QUESTION.

 

I apologize for the caps and the repetition, but given the state of the discourse, I was afraid that anything less emphatic would let us sink back into the mire. This is a post about the dysfunctional discussion, not about the underlying questions. It is entirely possible that our current fire safety laws around building access are onerous and that a good cost-benefit would find that we should phase some of them out. I can't comment on that because I haven't seen a good overview of the data and consensus expert opinion on the subject.  Unfortunately, neither have any of the other people who learned about the matter from a spate of articles from Slate, Lawyers, Guns and Money, New York Magazine, Slow Boring (Matt Yglesias), etc.

Instead, we get something like this from Eric Levitz.

Thus, the two-stair model comes with considerable architectural and economic costs. And it’s far from clear that multiple stairways meaningfully increase fire safety. Many European nations where single-stair designs are dominant have lower rates of fires than the U.S. Mandating sprinkler systems and other low-cost fire precautions would likely compensate for any diminution in safety resulting from a switch to a single-stair standard.

We get a couple of weaselly phrases like "far from clear" and "would likely compensate," and a standard other-countries-do-this-and-they're-fine arguments not dissimilar in form to the claims that smoking reduces heart disease in France (which is once again -- don't make me bring out the caps lock -- not to say that the comparison here is invalid, just that you have to be really careful with this kind of causal reasoning). It's also worth noting that the single link included doesn't point to actual data on the subject.

The closest I've seen to an actual discussion of the safety issues is from this widely cited Slate piece by Henry Grabar.

The specter of big structure fires—like the fire at London’s Grenfell Tower, the single-stair housing project whose defective façade panels caught fire in 2017, killing 71 people—is what reformers like Eliason and Speckert are up against. But building fires are much less common than they were when single-stair rules were codified, to the extent that most city dwellers roll their eyes at office fire drills and curse their hyperactive apartment smoke alarms. Data from the World Fire Statistics Centre show Canada, for example, has little to show for its two-story limit.

Putting aside the people complain about fire drills and smoke alarms so fire danger is not that big a deal (how many safety measures do people not complain about?), Grabar does link to actual data, which would be an improvement except for a few things: I get a 404 when I click on the link; it refers to Canada housing which is a highly problematic outlier (hell, I'm a YIMBY when it comes to Canada); the link isn't to the World Fire Statistics Centre. Instead it goes to the site of Speckert who seems to have built his career around advocating for this one issue and who presumably is picking data points that support his position; You might get the impression that the organization that puts out this data is on the same page as Grabar.

On that last point. I'm not sure which center/centre we're talking about (Speckert's site isn't much help). It could be this apparently dormant report from the Geneva Association, an insurance industry group, or it could be from the International Technical Committee for the Prevention and Extinction of Fire (French: Comité Technique International de Prévention et d'Extinction du Feu - CTIF). GA doesn't appear to have said anything about the single stairway question but CTIF has, and if you search their site, here's what you'll find.

One single staircase for a 570 feet building 

Concerns are recently rising in the UK about high rise and mid rise building safety in general, and it has been suggested rescue staircases should be designed differently based on what happened in the Grenfell Tower fire. Since an extra staircase would take away space which could be monetized as residential areas, it is thought that developers are resisting putting in an extra staircase, despite safety concerns.

The UK building code only requires one single staircase, even for a building as high as the Cuba Street Tower.

The international building code however - which is adopted by many countries and US states but not the UK - requires that, residential blocks need to be built with at least two staircases if the building is taller than four storeys.

One of the main safety concerns is that residents would need to use the same staircase as the firefighters would be using during a fire - which could lead to smoke inhalation for the residents trying to exit the building.

In an article in The Guardian on January 10, 2022,  UK fire experts criticized the design of the Cuba Street Tower, a new 51 storey apartment high rise close to Canary Wharf in London.  The apartment tower is planned to house more than 420 apartments and more than 650 bedrooms. 

 

London Fire Brigades writes Letter of Concern re: single staircase in a new 173 meter building

Despite that all apartments in the Cuba Street project above 11 meters are planned to be equipped with sprinklers, fire safety experts feel evacuations may still be needed in case of a fire.

An article in the Daily Mail, claimed on January 13 that the 428-flat tower block will be one of the tallest residential buildings in the UK. It will be 570 feet (173 meters) which is more than two-and-a-half times the height of Grenfell Tower. 

According to the same Daily Mail article, the London Fire Brigades have serious concerns about the single staircase and has addressed the city in a Letter of Concern:

"... We do not believe that sufficient justification has been provided for the tall single stair approach, nor do we agree that particular aspects of the design are compatible for such an approach. Furthermore, in our opinion there are insufficient facilities provided to support the safe egress for disabled occupants...".   

Arnold Tarling, a chartered surveyor and fire safety expert, said in the Guardian article: “It is utter madness that this is still allowed.”

Tarling allegedly has recently inspected a newly built apartment tower in the same area of London and discovered serious failings that would mean residents might not be safe to stay in their flats in a fire. 

He then referenced the recent deadly Bronx fire, where many casualties could have been avoided had fire escapes been installed in the building.  The worst-case of having only a single staircase would be  “another Bronx fire, another Grenfell, or another Lakanal type fire”, he said.

 Despite the impression you'd get from Slate, LGM, New York Magazine, Slow Boring, and pretty much every pundit who weighs in here, the safety of these proposals is still very much up for debate and actual fire fighters have landed firmly on the worried side.

The entire housing discussion- - if we can even call it that (discussion implies an exchange of ideas)-- has become so overwrought and distorted that it no longer even vaguely reflects the actual questions that need to be discussed. In this case we have a dispute between developers and architects on one hand and fire safety experts in First Responders on the other over the issue of safety regulations. It is enormously telling that Grabar without supporting arguments or even comment puts the reformers on the side of the first group. Historically, this is not the way this breaks.

None of this is to say, or even to suggest, that the single stairway people are wrong. There are any number of examples of excessive and onerous regulations with unexpected consequences that need to be revised or removed entirely. In case you missed the first six lines of my post, I don't have a position on that, but I do know that what we're getting from the pundits is not helping us make an informed decision.

 For an example of a much better example of an argument for fewer safety regulations, I send you to Mitchell and Webb.


Monday, April 17, 2023

Twelve years ago at the blog -- game nerds taking it to eleven

Friday, April 29, 2011

Weekend Gaming -- perfecting the imperfect

[disclaimer -- I've only field tested the first of these, so I can't guarantee that all of the variations will play smoothly. On the bright side, there ought to be plenty of room for improvement. As with all discussions of game variants, you should probably assume that countless people have already come up with any idea presented here.]

When the subject of perfect information games comes up, you probably think of chess, checkers, go, possibly Othello/Reversi and, if you're really into board games, something obscure like Agon. When you think of games of imperfect information, the first things that come to mind are probably probably card games like poker or a board game with dice-determined moves like backgammon and, if you're of a nostalgic bent, dominoes.

We can always make a perfect game imperfect by adding a random element or some other form of hidden information. In the chess variant Kriegspiel, you don't know where your opponent's pieces are until you bump into them. The game was originally played with three boards and a referee but the advent of personal computing has greatly simplified the process.

For a less elaborate version of imperfect chess, try adding a die-roll condition to certain moves. For example, if you attempt to capture and roll a four or better, the capture is allowed, if you roll a two or a three, you return the pieces to were they were before the capture (in essence losing a turn) and if you roll a one, you lose the attacking piece. Even a fairly simple variant such as this can raise interesting strategic questions.

But what about going the other way? Can we modify the rules of familiar games of chance so that they become games of perfect information? As far as I can tell the answer is yes, usually by making them games of resource allocation.

I first tried playing around with perfecting games because I'd started playing dominoes with a bluesman friend of mine (which is a bit like playing cards with a man named Doc). In an attempt to level the odds, I suggested playing the game with all the dominoes face up. We would take turns picking the dominoes we wanted until all were selected then would play the game using the regular rules. (We didn't bother with scoring -- whoever went out first won -- but if you want a more traditional system of scoring, you'd probably want to base it on the number of dominoes left in the loser's hand)

I learned two things from this experiment: first, a bluesman can beat you at dominoes no matter how you jigger the rules; and second, dominoes with perfect information plays a great deal like the standard version.

Sadly dominoes is not played as widely as it once was but you can try something similar with dice games like backgammon. Here's one version.

Print the following repeatedly on a sheet of paper:

Each player gets as many sheets as needed. When it's your turn you choose a number, cross it out of the inverted pyramid then move your piece that many spaces. Once you've crossed out a number you can't use it again until you've crossed out all of the other numbers in the pyramid. Obviously this means you'll want to avoid situations like having a large number of pieces two or three spaces from home.

If and when you cross off all of the numbers in one pyramid you start on the next. There's no limit to the number of pyramids you can go through. Other than that the rules are basically the same as those of regular backgammon except for a couple of modifications:

You can't land on the penultimate triangle (you'd need a one to get home and there are no ones in this variant);

If all your possible moves are blocked, you get to cross off two numbers instead of one (this discourages overly defensive play).

I haven't had a chance to field test this one, but it should be playable and serve as at least a starting point (let me know if you come up with something better). The same inverted pyramid sheet should be suitable for other dice based board games like parcheesi and maybe even Monopoly (though I'd have to give that one some thought).

I had meant to close with a perfected variant of poker but working out the rules is taking a bit longer than I expected. Maybe next week.

In the meantime, any ideas, improvement, additions?

Friday, April 14, 2023

Watching ChatGPT explain a reasoning problem is like watching someone who learned the language phonetically take an improv class

When trying to keep your head above the hype and bullshit surrounding large language models, here's a handy trick. Ask yourself how many people have answered exactly the same question in print over the years. If the answer is a large number, it would not be surprising if the LLM approach of looking at the likelihood of certain words and phrases appearing with other words and phrases actually produced something that seemed to show comprehension and awareness. 

For example, if you asked ChatGPT to prove some familiar theorem, there's a good chance that the algorithm would stitch together an acceptable answer out of the thousands of explanations out there. 

 


It may look impressive, but it no more demonstrates an understanding of mathematics than reciting a passage from Goethe learned phonetically demonstrates a mastery of German grammar and syntax.

Here's what it looks like when an LLM encounters a new problem.







It is not just that the answer is wrong or that the explanation is wrong; it's that they seem unaware not just of the problem, but of each other. The answer at the top is different than the answer at the bottom and neither matches the steps given. 

We need to start thinking of these systems not as generative AI but as regurgative. In some cases that's good enough, niche applications like generating unimportant boilerplate text or writing code snippets, but even in those limited area, it can only function where humans have not only explored a question, but have done so in such exhaustive detail that the algorithm can autocomplete its way to a useful response.