My old employer used to send me regular care packages: comfortable sweatpants, fancy candles, 1000-piece puzzles, and the like. It wanted our homes to feel as homely as possible, particularly during the pandemic, when there was so little to do outside. One of the packages consisted of three miniature plants. They were contained in pots no larger than one or two inches in diameter. I remember reading the names in the small informational booklet provided, and then promptly forgetting them after throwing it away. One looked like a grass, tall and spindly and blade-like, with several wooden rods inserted vertically into the soil to provide support. One was a short stubby plant, with small teardrop leaves, green and ribbed on the top, and deep red on the underside. It occasionally sent up slender flowers, soft and fuzzy to the touch, although it had no partner to mate with. And one, I later discovered, was a pilea, or a Chinese money plant. It grew so quickly under my care that I deceived myself into thinking I had a green thumb (It turns out pileas are “beginner plants” because they are so easy to care for.)

Continue reading “Seeds”


Hang ‘em with it


Dana Bash: You’re an outspoken opponent of abortion rights, but are you at all concerned that this [mifepristone ruling] sets a dangerous precedent that any single judge can just overrule scientific agencies as they see fit?

Rep. Tony Gonzales: Dana, thank you for having me, and happy Easter from San Antonio, and special shout-out, happy birthday to Jackie, my four year-old, who is our Easter baby. You know, on this rule, I have 6 children. I’m a prolific pro-lifer, and I think it’s important we protect the sanctity of life. I believe in states’ rights. Here in Texas, we have a heart bill, uh, a heartbeat bill that was passed, and I think it’s important that states dictate their futures, and you have to have the courts uphold these. It’s very dangerous when you have the administration, the Biden Administration, coming out and saying they may not uphold a ruling? As an appropriator, on the House Republican side, I look at it as the House Republicans have the power of the purse, and if the administration wants to not live up to this ruling, then we’re gonna have a problem. And there may come a point where House Republicans, on the appropriations side, have to defund FDA programs that don’t make sense.

Dana Bash: You said that you want this to be states’ rights, but isn’t a federal judge saying on a national level, that a pill cannot be administered, the opposite of states’ rights?

Rep. Tony Gonzales: Well the states started this, the states had their ruling, and now the federal government is coming in and dictating theirs. I think it’s important that, we have to get back and allow our institutions to lead. We can’t undermine them when we don’t agree with things that are there, whether it is things on the state level. Look, I’m from Texas, we don’t have marijuana here. Marijuana is in California, and other places. If those are the kinds of things your community wants, then work it through the state, work it through the federal level, but we have to uphold our institutions. It’s dangerous when we erode them.

Dana Bash: I want to move on, but I just want to point one important thing out, which is that mifepristone isn’t just used for abortion, it’s also frequently prescribed for women experiencing a miscarriage. By some estimates 1 million women miscarry every single year. So, are they just on their own, if this ruling is upheld?

No, I think it’s important that we take care of women. It’s important that we have real discussions on women’s healthcare and get off the abortion, you know, the abortion conversation. Women have a whole lot more other issues than just abortion, and let’s have those real conversations, and let’s talk about the other things that are happening in this world. You know, I’ve got a picture of, uh, Emily, uh Amelia, and Maria. They recently passed away due to a smuggler in my district. What does that mean? That means that there are all these other things happening in the world, especially in my district. You’ve got a district that’s been turned upside down, due to this border crisis, there’s everyday people that are impacted on this crisis, to include the Tambungas.

Dana Bash: Well, both things can be true. Everyday people can be affected by all of these issues facing Americans.

Continue reading “Hang ‘em with it”

Brother Rod


You might remember Rod Dreher from a post I wrote almost 6 years ago now. (Where did the time go?) It was titled “The Sociopath”, and in it I recounted what a gross, awful, and bigoted person he was. (By the way, he still is.)

Dreher has been a blogger at The American Conservative since 2008. During that time, in addition to being gross, awful, and bigoted, he has also been weird. Very weird. Dreher’s writing exhibits a prurience, particularly for gay sex, that is often associated with deeply repressed or closeted conservatives. (I’m not suggesting that Dreher himself is closeted; there are plenty of straight homophobes.)

Continue reading “Brother Rod”

We are ruled by the worst people, part infinity

Image from iOS

I survived a layoff this week. This is the third round of layoffs I’ve witnessed in three years (my previous employer was shut down a few months ago, which would have been a fourth).

Layoffs are strange. For one, it’s rarely clear why certain employees were chosen. One might imagine that “performance” is an important part of the decision, but this is typically not the case. At my company, high performers were let go along with low performers. People who were recently promoted were laid off, as were recent hires, who presumably haven’t even gotten used to the day-to-day responsibilities. My annual performance review hadn’t even been finalized, which raises the question of whether my performance was considered at all.

(This is not as bad as it sounds. Even if using performance as a criterion for layoffs might sound meritocratic, and even if some of my colleagues are incompetent, lazy, or both, there is no guarantee that the system could distinguish between a good employee and a bad one. As one of my colleagues remarked, even merely giving a colleague praise and criticism in a peer review would be fraught with danger if that review could subsequently be used in a layoff decision.)

Continue reading “We are ruled by the worst people, part infinity”



In June, as the sun set on Dublin, Ohio, a well-to-do suburb of Columbus, several dozen people dressed in golf shirts and floral shifts filed into a small auditorium to listen to a talk by a new neighbor. Vivek Ramaswamy, a thirty-seven-year-old entrepreneur, had settled in the area with his wife and toddler son after making a large fortune as the founder of a biotech company. Now, thanks to dozens of appearances on Fox News to criticize “cultural totalitarianism” enforced by liberal élites, he was closing in on fame as a conservative pundit. In the past year, he had cast aspersions on Black Lives Matter and “the death of merit”; mask mandates and U.S.-border protection; public-school curricula and the actor Jussie Smollett. All the flame-throwing had established him, in the words of one anchor, as the network’s “woke and cancel-culture guru.”

Continue reading “Anti-woke”

Human after all


I work at a music streaming company, and we spend a lot of time thinking about “recommendations”. At first glance, the importance of recommendations might not be obvious. Naively, we might imagine that a music streaming service operates like a library. When I go to a library, I typically have in mind the books I want to get. These have been recommended to me by friends or experts, or by anonymous reviews from faraway strangers. The function of the library, then, is just to provide a service to locate my desired book in the stacks. A music streaming platform, by analogy, might simply be a search bar connected to a play button: one in which, after I type “Lose Yourself”, Eminem starts rapping.

But no for-profit technology has ever been content with its own modest existence, and music streaming is no different. To some extent this makes sense. Listeners might want more out of a streaming service than literal streaming. They might want ways to organize their music into playlists, to share music with their friends, or to learn about new releases or upcoming concerts from their favorite artists. But they might want also someone else to tell them what to listen to next. Music is supposed to be a respite from hard work, not hard work itself.

One of the simplest recommendation algorithms, and likely the one taught first in most machine learning courses, is known as “collaborative filtering”. The basic idea is deceptively simple: similar users will listen similar pieces of content. User tastes are assumed to be stable over time. And, therefore, users will be most likely to listen to recommendations of songs that are similar to ones they have listened to before, or to ones that users similar to them have listened to before.

Unfortunately, we usually lack explicit measurements of either “user similarity” or “content similarity”. Instead, these similarities are inferred from past listening data. I won’t go through the math, but the algorithm learns a vector that represents a user, and one that represents a piece of content (in this case, a song). Crucially, these vectors are embedded in the same space, so that the distance between a user and a song vector can be interpreted as the “similarity”, in some sense, of the user’s taste to the song’s musical and cultural qualities. Songs close to a user in vector space are, presumably, better recommendations for that user than songs far away.

It is not difficult to see how this can become a virtuous cycle, at least from the company’s perspective. Users listen to music on the platform, implicitly revealing their “taste vector”. This listening data feeds into the recommendation algorithm (”collaborative filtering”), and it recommends increasingly better content the more data it gets. Higher-quality recommendations drive more users to the platform, and these users, in turn, generate even more data. Rinse and repeat. It’s worth noting that this cycle exists for most tech companies. In fact, it might even be taken as the definition of a tech company: i.e., a tech company is one whose algorithms grow more effective as their data expands, and whose data expands as their algorithms grow more effective. (See, for example, Shoshana Zuboff’s “The Age of Surveillance Capitalism”.)

The story is not quite so simple, though. Funnily enough, my employer largely does not use collaborative filtering at all, or at least not in its simplest form. Instead, user-created playlists are used to derive similarities. A song is taken to be similar to another song if users tend to put them together on playlists that they create. Users are similar to other users if the songs they listen to are similar. The playlist that you create for your own enjoyment has thus become the fundamental building block of the algorithm, and of its conception of your “taste”.

The problem of recommendation, whether in music or not, is largely a “metadata” problem. Songs have metadata, like artist names, release dates, albums, genres, Pitchfork ratings, Wikipedia entries, popularity scores, and the aforementioned “taste vectors”. Some of these pieces of metadata are largely trivial to obtain, but others are much harder. There is no easy way to categorize hundreds of millions of songs into “genres”, even assuming that “genre” is a stable and coherent concept to begin with. Turning from genres to even more contentious and vague concepts, like moods and aesthetics, the problem of assigning metadata becomes more challenging. Suppose you want to figure out if a song is “chill” or not, or if it’s the kind of music you’d play in a car with “windows down”. That information might be useful to a recommendation algorithm for a “Chill” or “Windows Down” playlist (or, more creepily, for a recommendation algorithm that knows you are in such a mood, or situation, and gives you exactly what you “want” without asking). There are typically two approaches. One is to rely on music experts to manually assign the “chill” metadata tag to songs they consider “chill”. Some such experts are my colleagues. The other is to use algorithms. The advantage of the algorithms is that they “scale” better than people do, even if their accuracy is markedly lower. Once again, you might find it surprising that the algorithms, at least at my employer, do not analyze the auditory qualities of a piece of music to determine its “chillness”. (Perhaps not so surprising, actually — chillness is as much a cultural phenomenon as an auditory one.) Instead, we once again rely on user playlists. What better signal exists that a song is “chill” than if one of our listeners adds it to their handcrafted “Chill 2022” playlist?

For all their purported sophistication, algorithms are largely, at least for now, simply regurgitations of our collective taste. An algorithm does not know what “indie pop” or “windows down” is, beyond the information supplied to it by music experts (some of my colleagues, although certainly not me) and non-experts (you). Given how fundamental user playlists are to music recommendation, I would be surprised if any machine recommendation, no matter how good, had not first been made, at least implicitly, by a human: one who chose to put together two songs — the one you knew already, and the one you just “discovered”. Any serendipity that the algorithm might convey is, perhaps disappointingly, simply a byproduct of statistical patterns in an almost incomprehensibly large dataset of our own creation. And so these moments of magic are, in my view, still fundamentally human, even if they are mediated by a machine. This is not to detract from the power of statistical summarization: the machine can remember and synthesize billions of connections between songs far better than any human can. But it’s worth remembering that each of these connections was first made by someone like you.

I was reminded of all of this while playing around with ChatGPT, Open AI’s most recent buzzworthy deep learning system. The basic idea of GPT-3, its underlying machine learning model, is shockingly similar to that for music recommendation. Instead of learning which songs co-occur in playlists, it learns which words co-occur in sentences and paragraphs. (One subtlety: order matters for words much more than it does for songs.)

Jay Caspian Kang writes about GPT-3,

If, for example, the word “parsimonious” appears within a sentence, a language model will assess that word, and all the words before it, and try to guess what should come next. Patterns require input: if your corpus of words only extends to, say, Jane Austen, then everything your model produces will sound like a nineteenth-century British novel.

What OpenAI did was feed the Internet through a language model; this then opened up the possibilities for imitation. “If you scale a language model to the Internet, you can regurgitate really interesting patterns,” Ben Recht, a friend of mine who is a professor of computer science at the University of California, Berkeley, said. “The Internet itself is just patterns—so much of what we do online is just knee-jerk, meme reactions to everything, which means that most of the responses to things on the Internet are fairly predictable. So this is just showing that.”

Caspian Kang later calls what ChatGPT does “a series of parlor tricks” performed by “a very precocious child”, and I tend to agree. But this childlike behavior can lead down darker paths.

One good example is Github Copilot, a controversial machine learning system that, as the name suggests, helps software developers write code more efficiently: like having a copilot at your side. In the demos I’ve seen, a developer simply has to write a function with a descriptive name and documentation string and Copilot will fill in the rest, in a programming language of that developer’s choice. Copilot is controversial because, once again, it was built on human labor — in this case, code uploaded to Github — and some of that code explicitly prohibits commercial use. (Github recently announced it will allow businesses to subscribe to Copilot, at the cost of $19/user/month.) Github has been denounced by organizations like The Free Software Foundation, and has been taken to court in a class action lawsuit, but, in the meantime, Microsoft (Github’s owner) is making millions of dollars from synthesizing (or to be tendentious, “plagiarizing”) the work of thousands of developers, without returning even a penny to any of them.

I want to close by making two points. First, even a “precocious child” still learns from humans. Every Wikipedia article, novel, Tumblr post, or Humble Politics Blog essay digested by ChatGPT, and every sorting algorithm or Javascript snippet fed into Copilot, was originally written by a human, in the same way that every song recommendation regurgitated by a music algorithm was originally ingested from a user playlist.

Second, much of what we do as humans is not terribly inventive or creative. And this isn’t true just of memes, or of Elon Musk’s humor. There are millions of jobs that involve sending rote emails, applying basic algorithms from Computer Science 101 to a slightly different application, or telling people to turn their computer off, and then back on again. Eventually, a system like ChatGPT will be capable of doing some or even most of these jobs, albeit likely with some assistance from an actual human. Whether this is good or bad is unclear: I think it largely depends on whether the benefits redound to labor, or to capital (although I know which one I’m betting on). What is disturbing, though, is that these systems created by synthesizing our labor, and built largely without our knowledge or direct consent, could end up replacing and even immiserating us.

The inflation story


March 2022

Q: There are a lot of economists skeptical that you can reduce inflation as much as you’ve penciled in without raising the unemployment rate. And I’m just wondering what are the mechanisms you see in reducing demand, outside of housing and autos, how do higher Fed rates reduce consumer demand, unless it’s through higher unemployment?

POWELL: If you take a look at today’s labor market, what you have is 1.7+ job openings for every unemployed person. So that’s a very, very tight labor market, tight to an unhealthy level I would say. So, in principle, let’s say that our tools work as you describe, and the idea is that we’re trying to better align demand and supply, let’s just say in the labor market. So if you were just moving down the number of job openings so they were more like 1 to 1, you would have less upward pressure on wages. You would have a lot less of a labor shortage, which is going on, pretty much across the economy, we’re hearing from companies that they can’t hire enough people, they’re having a hard time hiring. That’s really the thinking there, these are fairly well-understood channels, interest-sensitive and basically across the entire economy we would like to slow demand so that it’s better aligned with supply, give supply, at the same time, time to recover, and do a better alignment of supply and demand, and that, over time, should bring inflation down.

November 2022

Q: Do you see wages being a significant driver of inflation?

POWELL: You know, I think wages have an affect on inflation and inflation has an affect on wages. That’s always been the case. The question is, is that really elevated right now? I don’t think so. I don’t think wages are the principal story of why prices are going up. I don’t think that. I also don’t think we see a wage price spiral. But, again, it’s not something you can — once you see it, you’re in trouble. We don’t want to see it. We want wages to go up. We want them to go up at a level that’s sustainable and consistent with 2% inflation. We do think that given the data that we have this labor market can soften without having to soften as much as history would indicate through the unemployment channel. It can soften through job openings declining. We think there’s room for that. We won’t know that. That will be discovered empirically.

Q: If I could follow up on that. Thank you. The Fed has acknowledged in the past that the tools that you have don’t affect things like energy and food prices that stem from some of those conflicts overseas. And they’re some of the biggest pain points for consumer. As you pursue the current path you’ve outlined, is there a risk that some of those prices simply don’t come down?

POWELL: We don’t directly affect, for the most part, the food and energy prices, but it affects them at the margin. The thing about the United States is we also have strong — in many other jurisdictions, the principal problem really is energy. In the United States we have a demand issue. We have an imbalance between demand and supply in many parts of the economy. Our tools are well suited to work on that problem. That’s what we’re doing. You’re right. The price of oil is set globally. It’s not something we can affect. I think by the actions that we take, though, we help keep longer-term inflation expectations anchored and keep the public believing in 2% inflation by the things we do even at times when energy is part of the story of why inflation is high.

Continue reading “The inflation story”



About a year ago I remember being bored and home-bound and tired of my own cooking. There’s nothing I find as delicious as carbs, and I stumbled onto a site called “The Brot Box”. The concept is quite clever. The company makes loaves of bread in the German style, without enriching the flour or adding sugar as many American breads do. The loaves are partially baked, in Germany, and then shipped to the U.S. in refrigerated packaging. They can be kept in the freezer for many months, and, when you want to eat fresh bread, you simply thaw a loaf, heat your oven, and finish the baking process. When I tried it, I was very impressed by the result; it reminded me of the bread I’d get in a European bakery before a morning hike.

I find myself almost irrationally bothered by waste these days — perhaps a byproduct of living in one of the filthiest cities in the Western world—, and the waste associated with my Brot Box compelled me not to get a second box. The concept seems faintly ridiculous, when you think about it. You are paying for a large box full of several pounds of food to be shipped across an ocean at low enough temperatures that the food doesn’t spoil. This in turn requires even more weight in the form of icepacks and insulation for those icepacks. The makers of The Brot Box have, to their credit, tried to make this as eco-friendly as possible. For example, the plastic packaging material is dissolvable in water and won’t clog up a landfill. But even if the box scores high on biodegradability, it remains a large, and, even worse, completely unnecessary, waste of petroleum and carbon and energy.

Continue reading “Waste”

How are people paid to do this?


How very happy it makes me to be writing to you all! I am wrapping up my second week in what I’ve come to call my Second Season here…For those of you who don’t know me, I am the new…product lead, and I’m returning…after a 1 year stint…On my mind this week is how Homecomings can also be new beginnings.

I’ve had many very happy reunions over the past two weeks and have spent time refreshing my brain on how the organization works and what we’re trying to achieve. But I’ve been struck these past two (whirlwind, fun, amazing, firehose-y) weeks not by how familiar [this] feels, but by how much it feels new and different.

I’ve come to realize over the past two weeks that, just like the second seasons of our favorite podcasts or TV shows, my Second Season is going to be full of plot twists, new characters, new journeys and new problems to solve. And I’ll get to do it with some old friends and many new ones I make along the way.

I am focused on coming up to speed as quickly as I can and becoming as helpful as possible as quickly as possible. As I mentioned, there’s a lot, so bear with me. My door is open, so please — come find me. Tell me what’s working and we’ll brainstorm how to scale it. Tell me what’s not working, and we’ll do some solutioning together. Tell me your wild and crazy product ideas. Chances are, [this] is the team of people who can make that dream come true.

Looking forward to rolling up my sleeves with you and delivering outsized…value to our users… Let’s have some fun!

I work at a big company now, after several years of working at smaller ones. It’s striking to me how many people at my current job get paid to do, well, bullshit. I’ve written before about bullshit jobs, and I know I should not find this phenomenon shocking, but perhaps it’s different now that it is so intimate and immediate.

Continue reading “How are people paid to do this?”

It’ll get worse before it doesn’t get better


I went on another trip recently and, once again, had the fortune to spend it with “libertarians”.

The libertarian attitude towards issues like policing, trans rights, and abortion should be close, if not identical, to that of the left. However, I found myself instead having a sort of meta-debate: not that much of American policing is brutal and ineffectual, or that trans people should be allowed to live how they want to — which is the essence of libertarianism, I had thought? — but instead that Democrats are politically foolish for agreeing with these stances. (Meanwhile I wondered, what would the ideology that got 1.2% of the vote in the last election know about political strategy?)

Libertarianism, it seems, has transmuted into anti-leftism, or a strange centrist contrarianism. In fact, one of my traveling partners admitted that he was sympathetic to the left a decade or two ago when “the right was in power” but now that the situation had flipped, he viewed “woke” ideology as the greater peril.

Continue reading “It’ll get worse before it doesn’t get better”