Tuesday, February 28, 2023

From the large language model joke book

Smith College economist James Miller has a great example of a ChatGPT answer that looks pretty good until you drill down.

 


 

Monday, February 27, 2023

A modest defense of YIMBY

This is Joseph

Some of the best discourses on the blog are cases where Mark and I disagree. I need to give him a big win on the last iteration, on the question of whether it was worth pushing the Republicans on social security. Part of it was that I underestimated just how clever of a politician Joe Biden was and how he could turn the discourse into a political trap.

So let me see if I can do better today. YIMBY has a terrible reputation for focusing on the petty. But the situation, in housing, for Ontario is . . . amazing. Immigration is way up


The number of new houses that needs to be built to compensate for the years of bad policy is epic:



One housing unit per minute. At two hours per approval (low end estimate) then the 10 unit pathway to would require 10 cities running approval committees 24/7, with no hitches or requests for more information (the need for new housing suggests the need to build one unit every minute in the province of Ontario). How is it going


So the province is falling completely short of the goal, that is also completely inadequate to make up for the bad policy of the past. At a time when the population is rapidly increasing (e.g., see the student visa situation). 

Now, let us look at an example in Guelph


The fight is over 23 stories or the 18 which is currently zoned. The comments are mad -- questions of whether the housing is affordable. Remember, the province needs to double the current record of housing (in the 1970's) just to stay standing still. What is the letter writer concerned about?
Regarding the highrise at the Wellington lot. Nothing being built should be more than five storeys tall.

Creating wind tunnels and blocking out the sun and sky is a Toronto answer to keep developers happy.

Low rise is the best answer going forward. Community is important.

Tornadoes and power outages are more of a problem when the buildings are too high. Think, what then.
So we should just remove the idea of a high rise building?  

YIMBY seems to pick petty fights, but look at this from the big picture. In the case where things need to move quickly, every single project gets held up on the pettiest of things. London, Ontario:
Ted Donaldson, a nearby resident and opponent of the townhome plan, said planning staff in the report appeared “hellbent on shoehorning” the project in against the wishes of 71 neighbours.

“Infill is essential. Nobody here sees another single family home being built on this lot,” Donaldson told the committee. “Infill must complement what is already there. It’s like a sculptor adding finishing touches to an already great neighbourhood. Infill is not a sledgehammer.”

Critics of the project expressed concerns about damage to trees, potential drainage issues on the lot, limited parking and garbage collection at the site.

What are they hoping for?

 “Five of these units, set back from the road, I could definitely support,” she said, adding she can’t vote against the proposal at the committee because she is not a member.

Now, it is true that this particular project is not going to materially effect the housing crisis, should it end up reduced to 5 units instead of ten. But the real goal is at the end:

Opponents of the project say they are not opposed to development on the property, but believe a 10-unit townhouse cluster is out-of-synch with the character of the neighbourhood.

Critics also see the potential rezoning as a warning to other London homeowners, who could see similar high-density infill projects crop up in subdivisions filled with single detached homes.

Basically, there is no way to grow the population of the province and not change the character of the cities. between 2011 and 2021, Ontario grew from 12.8M to 14.2M, which is 1.4M new citizens in 10 years. If this high rate of growth continues, which is the federal policy, then the idea that density can be opposed will require entirely new cities to be created. 

The other nasty secret, is that single family detached housing adds value to a neighborhood. It is a popular type of housing, Those neighborhoods that resist density reap a significant financial benefit and if every small project needs to go through a political fight then we'll never manage to end the housing crisis except in an explosion. It is true that not every new house can help with affordability, but a dramatic shortfall, in the face of an accelerated population boom, is sure not going to result in affordable housing, either. At some point, massively restricted supply has to have a market effect. 

Friday, February 24, 2023

When watching YouTube can make smarter

I'm not making a general claim. Most of the videos you'll find on the platform won't do anything for your intellect and more than a few will actually make you dumber (and frequently poorer). Nonetheless, there are some bright spots, channels that consistently leave you not only entertained, but better informed and perhaps more thoughtful.

Here are a few to start your weekend.

Common Sense Skeptic -- CSS is best known for their take-downs of Elon Musk's more outrageous claims (colonizing Mars, solving the labor crisis with humanoid robots. etc.), but they also make make time to debunk a wide range of scams by actually running the numbers and citing mountains of relevant research,






LegalEagle -- Devin James Stone was a lawyer before he was a YouTuber and he is one of the best at making legal issues understandable without oversimplifying them. He also knows when to ask for help, often brings in guest lawyers who specialize in that day's topic.




Brick Experiment Channel -- I love real engineering and this is one of my favorite channels in the field.






David Mitchell's Soapbox -- Funny and thoughtful essays about everyday subjects like social conventions, language, and, in this case, advertising.




Thursday, February 23, 2023

Senator Rick Scott and Simple Solutions

This is Joseph.

One of the most pernicious mistakes that I see if the common fallacy that complex problems can be solved by a simple solution. This is often accompanied by the conjecture that the people in power just do not have the willpower to implement the obvious solution. I see it a lot with the national deficit -- just spend less without grappling with the huge choices and constituencies that are involved.

Enter Rick Scott.

All federal legislation sunsets in 5 years. If a law is worth keeping, Congress can pass it again. [Bold added by me]

 It has been replaced by:

All federal legislation sunsets in 5 years, with specific exceptions of Social Security, Medicare, national security, veterans benefits, and other essential services. If a law is worth keeping, Congress can pass it again. Note to President Biden, Sen. Schumer, and Sen. McConnell — As you know, this was never intended to apply to Social Security, Medicare, or the US Navy [Bold in original]

I think this change illustrates two things. 

One, upon close inspection all sorts of bad examples appear that look alarming. The example of the Navy jumps out immediately, but the US Navy is a construct of legislation and the idea that it would sunset every 5 years seems bleak. All you get is a ton of work constantly renewing legislation. including this one (unless it becomes constitutional). Other fun questions arise about all sorts of foundational laws. For example, the supreme court is mentioned in the constitution but all of the rests of the courts are established by legislation

Article III of the Constitution, which establishes the Judicial Branch, leaves Congress significant discretion to determine the shape and structure of the federal judiciary. Even the number of Supreme Court Justices is left to Congress — at times there have been as few as six, while the current number (nine, with one Chief Justice and eight Associate Justices) has only been in place since 1869. The Constitution also grants Congress the power to establish courts inferior to the Supreme Court, and to that end Congress has established the United States district courts, which try most federal cases, and 13 United States courts of appeals, which review appealed district court cases.

For example, the size of the supreme court comes from the Judiciary act of 1869, which I do not see in the intended list of essential services. Reading the constitution, I see very few details and no support for the lower courts independent of legislation. So could the federal court system simply vanish because congress got distracted or gridlocked on another matter? One presumes that serious answers to these questions need to be thought about in advance. 

One also wonders about the 1790 residence act. Or the 1952 Immigration and Nationality Act. Or the 1878 Posse Comitatus act. This is really shooting fish in a barrel. There are hugely important laws that form the basis of civil society, which is why I bolded legislation in the first quote above. 

Finally, this is no longer simple. Now, every law has to be categorized as "sunset" or "doesn't sunset", a herculean task which could be equally accomplished by just having a plan to sunset unnecessary rules. No congress can bind a predecessor. But these sort of mass sunset plans tend to go poorly even when the laws in question are a small portion of the total.  Because current laws may have replaced other important laws and there are some quite unexpected interactions that occur. 

Now, old laws get struck down all of the time, although maybe this wasn't the example that Senator Scott was aiming for. But I think the real answer is that this document was rhetorical in nature, and not intended as a serious proposal. But I do think it illustrates the problems of "one simple solution" and the benefits of a careful engagement with the underlying issues. 

That said, I got this far without mentioning social security, so my editor will be annoyed. So let me say that I agree with Josh Marshall that the goal is to cut social security. I think that the coming demographic shift is unpleasant to deal with -- there is no way to keep benefits at current levels indefinitely and not raise additional revenue. I do think that the sums involved are a lot more modest than the doomsayers say, but that they are enough to cause some pain. This is a hard problem, which is why it has not been simply solved already. But fixing it requires a real discussion about trade-offs and not a simple idea.  

Wednesday, February 22, 2023

More SoCal Weather

 We're having quite a year.

Blizzard warning issued for SoCal mountains, with snow also expected at unusually low elevations

LOS ANGELES (KABC) -- A storm expected to be the coldest of the season is blowing into Southern California, bringing chilly rain and snow at low elevations.

The snow is expected to fall as low as 1,000 to 1,500 feet, meaning areas like Santa Clarita and lower-lying areas of the Inland Empire will see a rare coat of powder.

And at higher elevations, the National Weather Service has issued a blizzard warning starting Friday morning for the mountains of Ventura and Los Angeles counties. The service predicts from 2 to 5 feet of snow could accumulate in the mountains above 4,000 feet, falling even as heavy winds gust up to 75 mph.

Below that, at elevations of 2,000 to 4,000 feet, about 6-12 inches of snow are expected.

Visibility at that time is expected to be very low and travel is not advised through those areas. The blizzard warning is in effect from 4 a.m. Friday to 4 p.m. Saturday.

Passes like the Grapevine [I-5 from LA to the Central Valley and the Bay Area. -- MP] and the Cajon Pass are likely to also see dangerous driving conditions. Drivers are advised to bring chains and a full tank of gas and be prepared for difficult weather and road closures.

"They're expecting snow to drop as low as 1,000 feet," said Mark Bishoff with Caltrans. "The top of the Grapevine is a little over 4,000 feet, so they're expecting it to be impacted by snow." 

Just for a bit of context, the highest point in the city limits of LA is slightly over 5,000 feet. In the county, it's slightly over 10,000 feet.

Tuesday, February 21, 2023

Tuesday Tweets -- Hype and Glory

We start with a reminder that much of the world's trouble is caused by profoundly damaged people who desperately want attention.

Laura Jedeed has the disturbing details.
Project Veritas’ 2020 voter-fraud allegations are the organization’s bread and butter: likely the biggest reason for its $22 million haul that year. Yet O’Keefe uses the allegations solely to lead into his blood feud with The New York Times. His voiceover provides intricate legal arguments for the defamation lawsuit recently filed against the Times while O’Keefe himself — the real one, not the actor — dances onstage. The dancers in haute-couture newsprint dresses contort themselves as Lady Gaga replaces Jamiroquai and here we are, back where we began.

“You and me are like a bad romance,” Gaga sings, and I have to say, she’s not wrong. As the song reaches its crescendo, O’Keefe’s voiceover describes a 2021 encounter with the executive editor of The New York Times, who refused to acknowledge O’Keefe’s existence, which the real O’Keefe acts out onstage.

“In that moment, the muckraker had to come to grips with the fact that this supposed paragon of investigative journalism would never give him the time of day, and would never acknowledge his very humanity,” the voiceover says, referring to himself, as he does throughout his latest book, in the third person. The New York Times dancers claw at the real O’Keefe as he staggers to the front of the stage, heartbroken. “That small part of him that still hungered for recognition and acceptance from the ‘legitimate press’ — he once read The New York Times every morning — would never be satisfied.”

 

Segue to politics...


Worth noting that Trump is now going after DeSantis for his stand on Social Security and Medicare (a position shared by most of the governor's fellow Republican politicians but by few of their constituents).

All of this may be having an effect.

2022

Orlando CNN — Walking out to a slick hype video and tossing hats into a raucous crowd as he approached the microphone, Florida Gov. Ron DeSantis on Thursday demonstrated for the Conservative Political Action Conference the bravado and fighting attitude that has made him the most popular elected Republican in the country among conservatives.

2023


"Dems in disarray"



Just a reminder, the only clear policy difference between the two leading GOP candidates is that Trump has not gone anti-vax.


Putin and MAGA


Long thread.


And another.

Marshall reads the NYT for us.
Karl Bode also reads the NYT for us.


OpenAI may be the most overhyped company in living memory.


 

One more from Grady.



And taking us full circle to the  first sentence of the post...

Monday, February 20, 2023

Small update on Sinema

This is Joseph.

We've been interested in the Krysten Sinema story from a game theory perspective -- was it a brilliant move to go independent to avoid a Democratic primary race. Josh Marshall reported on how many DC insiders see this as a great move. Here at the blog, Mark and I were skeptical. 

But then I saw this article which pointed to actual polls. If these polls are remotely accurate then it is terrible news for this plan. Now, the pollster points out:
Ruben Gallego is leading all his potential competitors (a good sign for Gallego) but in no matchup does he exceed the 50% mark (a good sign for the GOP and Kyrsten Sinema).

Ruben Gallego being the front runner for the Democratic nomination, should the Democrats decide to run a candidate.  This under 50% is a true statement but the question here is whether it makes sense for Sinema to run. 


Looking below at favorability, Gallego is +6 with independents and +56 with democrats. Sinema is -6 with independents and -19 with democrats (only -4 with republicans, her strongest group). But the three possible replication candidates range from +21 to +30 with republicans, so it isn't like that is a natural base for her to poach important levels of support. She does beat all three possible republicans with independents, but Gallego does even better without losing support from his own party. (Doug Ducey is in the polls but not the favorability ratings). 

But these are completely intuitive results. The worst case for Gallego is running against a moderate republican (it is a purple state) and not having Sinema in the race. That said, most of the time she hurts the republican more. Look at Doug Ducey, who looks like Gallego's toughest fight at this stage. 

With Sinema: D 32, I 17, R 27, Undecided: 23

Without Sinema: D 38 (+6), R 34 (+7), Undecided 28 (+5)

There is a general pattern of Sinema possibly hurting the Republican candidate, stealing more support from them, more than the Democratic candidate. Now undecideds could break heavily R and create a path to victory for Doug Ducey or Blake Masters (even the weakest R candidate is currently outpolling Sinema). But what is the advantage to either side in not running a candidate?

Sinema votes for D priorities the majority of the time, including judges. Why would Rs not run a candidate with that pattern of favorability (Karrin Taylor-Robson is more popular with Ds than Krysten Sinema, an impressive feat). But the argument against Ruben Gallego running crucially depends on Sinema winning -- if the Rs could field a candidate that could pull in the D base (and, obviously the R base) then what is the benefit of losing while doing well with independents? Keep in the mind, many of the normal mechanisms that would bolster support for the Democratic candidate like "they won the primary" and "our candidate, right or wrong" are not going to work with an independent who specifically avoided obtaining this support. That the Republican candidate might poach support from the Democratic base against the independent candidate is a clear sign of how extraordinary this situation might be. 

Finally, the threat of going negative has to considered in context of whether you want this to become a common tactic to remove the D senate candidate. Consider:

Sinema’s allies say that Gallego will get tarred as too left wing and also hint darkly that they’ve got a load of oppo to use against what Palmeri oddly terms the “twice-married Gallego.” Either might be true. I have no idea. But neither makes Sinema more viable.

But any race could have an independent appear and start smearing a candidate. If you give into these threats then you'll be constantly abandoning races. The key is that Sinema has a -19 favorability among her own party in her own state. That is a basic political skill. 

To make the independent trick work, I suspect that you need to personally be very popular (so that stepping aside is a pretty clear win) as well as being a pretty reliable ally. Angus King is a sitting independent senator and is 62-28 (+34) in his state of Maine. Bernie Sanders has a favorability rating of 64 in his state of Vermont in 2020, down from an epic 80-17 (+63) in 2016. Sinema is 37-47 (-10) in her state of Arizona. Neither Sanders or King gets national headlines attacking D priorities like the minimum wage from the right

All of this to say that the polls just don't support this maneuver and her conduct as a senator is probably why they don't. It doesn't help matters that Mark Kelly has won as a D twice since Krysten Sinema's historic win, making Arizona a state with two D senators, at least until Krysten Sinema went independent. But that leaves open the question of why not try to have two again?

So, I agree with Josh Marshall. The path to victory here for an independent with these favorability ratings is very narrow indeed. 



Friday, February 17, 2023

One more rant on the libertarian idea of children

This is Joseph.

There has been a lot of angst about global fertility rates, recently. I always find it ironic when these complaints come from people with the resources to combine work and childcare. For example, Marissa Mayer famously had the office next to her turned into a nursery, not a typical strategy. Or Elon Musk, who has many children, has a different level of wealth for supporting them.

In the United States there is also a very distinct rural culture where a lot of what I am saying does not apply, as these costs are very different. There is a notable rural/urban fertility gap  in the US, and I am suspicious that the issues may be quite different in the rural context. Peter Zeihan seems to think so:


But let us put that aside for the moment and consider the urban context for children. In an urban context, children are a huge and expensive responsibility that a lot of the world is not designed for. Doubt me? Ever try to take a taxi with a 3-year old? How do you meet the legal mandate for a car seat? If you think the answer is to carry one around and just ask the Uber or taxi to let you install it for the trip . . . 

Or what about bad school planning? There are huge consequences if a child is unable to go to the local school and needs to be driven to a distant school at which they are assigned. This example is especially infuriating because the family needs to buy a car to do drop-off because there is no way to make bus service make sense. Now you might say that these things will happen but this is offsetting a large cost to parents. 

Look at the cost of childcare, while we are on the subject. Daycare in Seattle, for example, is $1480 per month. That means daycare (alone, not counting the cost of feeding or housing the children) would eat up an entire $42K a year salary. For the median salary in Seattle, 2 kids is 44% of after tax income in childcare costs. 

So why do I bring this up? Because, while I think the doomsayers are wrong, it isn't a great surprise that huge costs and low support are depressing fertility rates. And, whether we like it or not, children are the future of the society that we live in. Making it hard to have children is going to result in fewer of them. 

This seems to be a consequence of the libertarian idea of economic atomism, where the only way to see children is as a personal luxury consumption good. But I fear that ends up bringing us to a dark place. 


Thursday, February 16, 2023

If a TV show runs in the city and nobody sees it...

When I came across this article from New York Magazine (technically the Cut, but you know what I mean) on on Fleishman Is in Trouble, I had originally intended to do a post making fun of the many passages like this.

Since leaving New York, Beth has found herself in tears at least once a week. She makes $300,000 a year — more than she’s ever earned in her life — but she’s running out of minutes in the day to squeeze out more dollars. “How do I make the $700,000 that I’m going to need to send her to private school or do the renovation in the attic so I can turn it into the master suite so I can have a tub and so I can have one thing I enjoy in my life?” she says. Her takeaway from the show: “Both avenues are shit. You can stay in New York and climb, climb, climb and never get where you need to go and give yourself a nervous breakdown, or you can move to the suburbs and be like, Who the fuck are these pod people? Neither seems great. Is the secret to it all that we have to just choose a lane and embrace it?” 

The national press, particularly publications with "New York" somewhere in their name), never tire of telling us about the financial and emotional hardships faced by the bottom half of the top one percent. By the standards of the genre, the NYM piece lacks the hilarious budgeting assumptions explaining how a middle class couple can find it hard to scrape by on $300,000 or the stunning cluelessness of a Bret Stephens who thinks a couple in SF making $400,000 are lucky to manage a Camry, still it's hard to beat lines like "so I can have one thing I enjoy in my life." 

But as I started to read up on Fleishman, I started thinking this story might fit better with another long running thread.

The series has gotten a ton of coverage...


... which means (and I apologize for disillusioning some of our less worldly readers) Disney is spending a ton on PR. The streaming industry runs on hype and easy to promote awards bait play a big role.



Whenever you're reading about these shows, the first question you should ask is "how many people are actually watching. (The second question is "who owns the IP?"). It's often difficult to find out -- streaming services are secretive about these numbers -- but FlixPatrol is probably as good as we'll get. Here's their list of Hulu shows ranked by viewership for 2022.

For a sense of what is popular, here are the top 20. (check out number 5)

1.         Family Guy    
2.         P-Valley    
3.         The Kardashians
4.         Power Book IV: Force    
5.         General Hospital    
6.         Law & Order: Special Victims Unit     
7.         Bob's Burgers     
8.         Power Book III: Raising Kanan     
9.         House of the Dragon
10.       The Chi     
11.       Euphoria
12.       Power Book II: Ghost    
13.       9-1-1    
14.       Love Island     
15.       Only Murders in the Building
16.       Grey's Anatomy     
17.       Abbott Elementary     
18.       The Patient
19.       The Good Doctor    
20.       This Is Us     

 If you go down the list (or use control-F), you find FIiT at 97 out of 119.



Though we can't say exactly how many viewers it takes to get to position 97, we can be pretty sure it's a very small number by TV standards. You almost have to wonder... If you took all the people who wrote articles about Fleishman Is in Trouble, and all the people quoted in those pieces, is it possible you'd have a majority of people who actually watched the show?


Wednesday, February 15, 2023

Whatever happened to the Washington Post's "The Truthteller"? -- repost

 It debuted ten years ago to a great deal of hype (much of which has a definite 2023 feel), then, as far as I can tell, faded quietly away. Now that automating journalism is back in the news, it might be interesting to revisit what happened to the last next big thing in the field.

Monday, February 18, 2013

Yeah, I'm getting tired of these too

Once again I made the mistake of listening to On the Media. It started out reasonably benign with a standard but harmless piece on cyber-security, then went downhill with a smug piece on fact-checking before going off a cliff with this ddulite puff piece. Here's the blurb:
Late last month, The Washington Post debuted "The Truthteller," an application that it hopes will soon be able to fact-check politicians' speeches in real time using speech-to-text technology and a vast database of facts. Brooke talks to Cory Haik, The Washington Post's executive producer for digital news, about the app.
Of course, the speech-to-text and database problems are trivial next to the issues with processing natural language. To work at anywhere near the level discussed by Haik, the system would have to be considerably more advanced than IBM's Watson. Watson was designed to address short, free-standing questions following similar linguistic conventions and having clear, unambiguous answers.

This isn't meant to denigrate the team that developed Watson. Just the opposite. Interpreting natural language is extraordinarily difficult and solving even highly constrained problems is an impressive and important accomplishment. IBM has a lot to brag about.

The Washington Post currently has a beta up that apparently can sometimes spot strings that look like simple factual statements that lend themselves to automated comparison to a database. There's no reason to believe the app will ever move much beyond that, but the interviewer did believe...

without sign of suspicion...

immediately after a segment boasting about how carefully On the Media checks its facts.

To quote Snoopy, "the mind reels with sarcastic replies."

 

Tuesday, February 14, 2023

Supplemental material for Prof. Narita's class

In case you aren't familiar with the professor's research (from NYT via Gelman):

     A Yale Professor Suggested Mass Suicide for Old People in Japan. What Did He Mean?

    In interviews and public appearances, Yusuke Narita, an assistant professor of economics at Yale, has taken on the question of how to deal with the burdens of Japan’s rapidly aging society.

    “I feel like the only solution is pretty clear,” he said during one online news program in late 2021. “In the end, isn’t it mass suicide and mass ‘seppuku’ of the elderly?” Seppuku is an act of ritual disembowelment that was a code among dishonored samurai in the 19th century.

...

Some surveys in Japan have indicated that a majority of the public supports legalizing voluntary euthanasia. But Mr. Narita’s reference to a mandatory practice spooks ethicists.

Viewing of supplemental materials is, of course, required.




Monday, February 13, 2023

Dems in Disarray

Remember all of the "Trump is fading away" articles a while back? These think pieces combined wishful analytics with a curious belief that Trump would peacefully go along with being pushed aside (even though the nomination would at least partially avert or at least delay indictments).

I'm still not seeing it.







And it gets better.

From Alex Henderson:

Right-wingers ranging from the Daily Wire’s Ben Shapiro to Fox News’ Rupert Murdoch to firebrand author Ann Coulter are very bullish on Florida Gov. Ron DeSantis as an alternative to Donald Trump for the 2024 GOP presidential primary. But The Lincoln Project’s Rick Wilson, a Never Trump conservative and former Republican strategist, has a very different viewpoint. Wilson has predicted that ultimately, the GOP will "bend the knee" and give Trump the 2024 nomination after his attacks on the Florida governor grow increasingly vicious.

Although DeSantis has yet to officially declare a presential run, Trump’s attacks on him are — just as Wilson predicted — becoming nastier and more mean-spirited. In a Truth Social post, Trump accused DeSantis of "grooming high school girls with alcohol as a teacher." DeSantis’ supporters were quick to call Trump out, but in a column published by The Bulwark on February 9, Never Trump conservative Tim Miller (a former GOP strategist) accuses them of being "hypocrites" who can dish it out but can’t take it.

"For some MAGA observers," Miller observes, "there is one thing startlingly new-ish about the former president’s latest broadside. It’s his target: Florida Gov. Ron DeSantis, the prized stalking horse of the conservative infotainment industrial complex…. If you missed it, Trump’s latest attack featured a picture in which DeSantis is shown drinking with women who appear younger than him and are allegedly his students. One included the caption 'Here is Ron DeSanctimonious grooming high school girls with alcohol as a teacher.'"



In case you've forgotten.


 

The real victim here is Elon. Cheong and Catturd are two of his new besties. Now they're fighting and he's caught in the middle.




Friday, February 10, 2023

Checking in with Elon

If you were writing a satirical roman Γ  clef of the Twitter take-over, you'd throw this scene out for being to on the nose.

 ZoΓ« Schiffer, Casey Newton writing for Platformer.

For weeks now, Elon Musk has been preoccupied with worries about how many people are seeing his tweets. Last week, the Twitter CEO took his Twitter account private for a day to test whether that might boost the size of his audience. The move came after several prominent right-wing accounts that Musk interacts with complained that recent changes to Twitter had reduced their reach.

On Tuesday, Musk gathered a group of engineers and advisors into a room at Twitter’s headquarters looking for answers. Why are his engagement numbers tanking?

“This is ridiculous,” he said, according to multiple sources with direct knowledge of the meeting. “I have more than 100 million followers, and I’m only getting tens of thousands of impressions.”

One of the company’s two remaining principal engineers offered a possible explanation for Musk’s declining reach: just under a year after the Tesla CEO made his surprise offer to buy Twitter for $44 billion, public interest in his antics is waning.  

Employees showed Musk internal data regarding engagement with his account, along with a Google Trends chart. Last April, they told him, Musk was at “peak” popularity in search rankings, indicated by a score of “100.” Today, he’s at a score of nine. Engineers had previously investigated whether Musk’s reach had somehow been artificially restricted, but found no evidence that the algorithm was biased against him.

Musk did not take the news well. 

“You’re fired, you’re fired,” Musk told the engineer. (Platformer is withholding the engineer’s name in light of the harassment Musk has directed at former Twitter employees.)

 There are lots of other last-hour-of-Titanic details in the piece including outages, internal chaos and this sure sign of a healthy $44 billion company

Thursday, February 9, 2023

“And it created this confidence that you could always rely on just knowing ... the El NiΓ±o/La NiΓ±a phase, and that would be able to give you a very accurate prediction for precipitation, especially for California. And that is just not scientifically true.”

A timely example of how the nuances in predictive models have a way of getting lost.

Rong-Gong Lin II writing for the LA Times.

For decades, two climate patterns in the Pacific Ocean have loomed large in predicting weather in California and other parts of the globe. El NiΓ±o — a warming of sea-surface temperatures in the tropical Pacific — seemed synonymous with wet winters for Southern California, while La NiΓ±a was a heralder of drought.

But the would-be model didn’t hold up this winter. Despite La NiΓ±a’s presence, a robust series of 10 storms brought impressive precipitation across California, spurring floods and landslides, increasing reservoir levels and dumping eye-popping snowfall in the mountains.

The Sierra Nevada has a snowpack of 240% of average for the date, and 126% of where it should be by the start of April. San Francisco was drenched with more than 18 inches of rain since Christmas, posting its wettest 22-day period since 1862. Downtown Los Angeles has logged more than 13 inches of rain since October — more than 90% of its annual average of 14.25 inches.

Though winter isn’t over, and a renewed dry spell can’t be ruled out, the significant storms have defied expectations of a dry winter.

The forecast in October by the Climate Prediction Center, a division of the National Weather Service, indicated the odds were stacked against the Golden State: a rare third year of La NiΓ±a was expected. And California had already recorded its three driest years in the historical record.

The center’s seasonal forecast for December, January and February said there were equal chances of a dry or wet season in Northern California. But for Southern California, the agency reported there was a 33% to 50% chance of below-normal precipitation.

Taking the midpoint of that forecast — say, 40% — that meant there was a 35% probability of near-normal precipitation and a 25% chance of above-normal precipitation, said David DeWitt, director of the Climate Prediction Center.

“These probabilities are going to be relatively modest ... because that is the state of the science,” DeWitt said.

Those subtleties, however, tend to get less attention. Easier to understand was the bottom line, as a center’s statement noted: “The greatest chances for drier-than-average conditions are forecast in portions of California,” as well as other southern parts of the nation. 

...

It was that jolt that pushed scientists to figure out ways to predict the next El NiΓ±o. The failure to forecast the 1982–83 event led to the development of a range of tools that successfully predicted another El NiΓ±o in 1997–98, which came in at record strength.

There was “massive flooding over the West Coast, especially California. And it was well predicted,” DeWitt said. The damage in California was severe — with at least 17 deaths — and brought Los Angeles its wettest February on record.

“And then the next year, 1998–99, was a strong La NiΓ±a, and you saw exactly the opposite ... these very dry conditions,” DeWitt said.

“And that imprinted on a lot of people — including the scientific community — a couple of messages: one, that that was what you were always going to see with El NiΓ±o and La NiΓ±a, especially significant-strength ones; and that basically, this was a solved problem.

“And not one of those was ever true,” DeWitt added.

He remembers his predecessor at the Climate Prediction Center testifying to Congress about the upcoming 1997–98 El NiΓ±o and its predicted effects, a forecast that ended up being on the money. “And it created this confidence that you could always rely on just knowing ... the El NiΓ±o/La NiΓ±a phase, and that would be able to give you a very accurate prediction for precipitation, especially for California. And that is just not scientifically true.”


Wednesday, February 8, 2023

Twelve years ago at the blog --How to lie with statistics -- rare cinema history edition

 

Friday, February 18, 2011


From David Leonhardt (via DeLong):
When I read that John Boehner, the speaker of the House, had said that the federal government added 200,000 federal workers under President Obama, I wondered, “Really? Where?” I’m not aware of any major federal hiring initiatives since January 2009.

... It turns out that the 200,000 number is simply incorrect.

...

Second, Mr. Boehner was starting his clock in December 2008, the month before Mr. Obama became president. The Bureau of Labor Statistics conducts its monthly survey during the week that contains the 12th day of each month, so there is no reason to start the clock in December 2008 as opposed to January 2009. On Jan. 12, 2009, George W. Bush was still president.
To be more accurate there wasn't an honest reason. As Leonhardt points out later, 11,000 jobs were added by President Bush in his last month in office. Speaker Boehner was interval shopping, one of the most effective and time-honored methods of lying with statistics. (Given that, by Leonhardt's estimate, Boehner went from 57,000 actual jobs to a claim of 200,000, he used lots of effective and time-honored methods of lying with statistics.)

Interval shopping is based on the idea that if you can adjust the period being studied, you can make something look much better or worse than it actually is. For example, if you take one day off of the service record of the Titanic, it looks like a remarkably safe form of transportation.

The method also allows you to have a great deal of fun with denominators. You will often see people in positions of responsibility pointing to a period of growth that started just after a disastrous collapse and ends just before the next one. The worse that initial collapse was, the better your growth rate looks.

Interval shopping can be particularly effective when the groups being compared are at different stages of life. You can, for example, use it to argue that a product is less reliable than one that was introduced a couple of years later, not taking into account the difference in average ages, or you could 'prove' the mental inferiority of one immigrant group over another by comparing test scores, not taking into account the higher proportion of non-English-speaking first generation immigrants.

For beautiful example of egregious interval shopping, check out this excerpt from a rebuttal to Gore Vidal written by Peter Bogdanovich in the New York Review of Books:
Now I’m getting in a foul mood because I’m reading this sentence again: “The badness of so many of Orson Welles’s post-Mankiewicz films ought to be instructive.” That’s another of those glib, sweeping statements that play right into the reader’s lack of information and is written so as to presume a general critical atmosphere, which in this case is not just superficial, it is decidedly untrue, which makes it all the more offensive and irresponsible on Gore’s part. Almost everyone with any sense knows that Orson Welles is a great director and that Herman Mankiewicz was a talented hack,* but for the record, here is a list of the movies Orson Welles has directed since Citizen Kane:

The Magnificent Ambersons

The Stranger

The Lady from Shanghai

Macbeth

Othello

Mr. Arkadin (Confidential Report)

Touch of Evil

The Trial

Chimes at Midnight (Falstaff)

The Immortal Story

F for Fake

And these are all of Herman Mankiewicz’s post-Welles films:

Rise and Shine

Pride of the Yankees

Stand by for Action

Christmas Holiday

The Enchanted Cottage

The Spanish Main

A Woman’s Secret

The Pride of St. Louis

One of the surest signs of interval shopping is the arbitrary start point, but the key to making it work is finding an arbitrary point that doesn't look arbitrary. Here Bogdanovich is able to make use of a sloppy writing by Vidal. The phrase "post-Mankiewicz" implies that there is some special significance to these films coming after Citizen Kane. If Vidal were comparing Welles' post-Mankiewicz films to his pre-Mankiewicz films (which he obviously isn't), or if he were arguing that Welles was changed by working with Mankiewicz (which seems unlikely, though I'd need to get behind the paywall to be sure), then the wording would have been appropriate. Here, though, we simply have Vidal saying "post-Mankiewicz" when he means "non-Mankiewicz."

This small bit of imprecision on Vidal's part gives Bogdanovich the opportunity to use Kane as the start point for his interval (and Peter Bogdanovich has never been one to pass up on opportunity). When comparing careers you would normally look at entire careers. This interval includes all of Welles' films and less than half of Mankiewicz's.

To make matters worse, the intervals aren't even close to the same length for the two men. Mankiewicz drank himself to death in 1953. Welles died in 1985 (the last film on Bogdanovich's list was released in 1974).

More importantly, though, this list includes all of Orson Welles' career as a director barring some shorts and TV work, while it leaves out most of Mankiewicz's major accomplishments as a writer and producer. Even in his final, declining, alcohol-soaked years, Mankiewicz still managed a good picture or two, but a list of films that he wrote or produced before Kane would include Dinner at Eight, Million Dollar Legs (with W.C. Fields) and three out of four of the Marx Brothers' best movies Monkey Business, Horse Feathers and Duck Soup.

And when you leave out Duck Soup, that's just going too far.



*Bogdanovich's senseless group here include Mankiewicz collaborators and admirers such as Alexander Woollcott, Heywood Broun, Dorothy Parker, Robert E. Sherwood, George S. Kaufman, Marc Connelly and Nunnally Johnson, but that's a topic for another post and perhaps another blog.

Update: The conversation continues here and here.