It’s a common and (on its face) rhetorical move: take something that’s happening now and map it onto the past. Better yet, take something atrocious that’s happening now and show how it maps onto something atrocious in the past, ideally affecting the very people who are now supporting the atrocities. “See?” this trope says: “what you’re doing to other people is exactly what was done to you.”

That’s the basic structure of “resistance genealogy,” as seen in clashes over immigration. “Tomi Lahren’s great-great-grandfather forged citizenship papers; Mike Pence’s family benefited from “chain migration”; James Woods’ ancestors fled famine and moved to Britain as refugees,” etc.

Rebecca Onion argues, convincingly, that this doesn’t work:

The chasm between the life and experiences of a white American, even one who’s descended from desperate immigrants of decades past, and the life of this Honduran mother is the entire point of racist anti-immigration thought. Diminishment of the human qualities of entering immigrants (“unskilled” and “unmodern” immigrants coming from “shithole” countries) reinforces the distance between the two. People who support the Trump administration’s immigration policies want fewer Honduran mothers and their 18-month-olds to enter the country. If you start from this position, nothing you hear about illiterate Germans coming to the United States in the 19th century will change your mind.

Besides underestimating racism, it flattens out history, and assumes that if people only knew more about patterns of historical racism, they might be convinced or at least shamed into changing how they talk about it. Everything we’ve seen suggests that isn’t the case.

I’m going to take this one step further and say this is a weakness in most resorts to historical and political analogies deployed as a tool to understand or persuade people about the present.

For example, consider Donald Trump saying, regarding immigrants trying to enter the United States, “these aren’t people, these are animals.” This is a disgusting thing to say and way to think — and not just because German Nazis and Rwandan perpetrators of genocide used similar language in a different context, and regardless of whether he was using it to refer to immigrants in general or members of a specific gang. It’s bad, it’s racist, it’s shitty, and you really don’t need the added leverage of the historical analogy in order to see why. But that leverage is tempting, because it shows off how much we know, it underlines the stakes, and it converts bad into ultra-bad.

This hurts me to say, because I love history and analogies both. But there’s a limit to how much they can tell us and how well they work. And playing “gotcha!” is usually well beyond the limits of both.

Like many of you, I read the news of a single person killing at least 10 people in Santa Fe, Texas today. While this is an outrageous and horrifying event, it isn’t surprising or shocking in any way in a country where more than 33,000 people die from gun violence each year.

America is a stuck in a Groundhog Day loop of gun violence. We’ll keep waking up, stuck in the same reality of oppression, carnage, and ruined lives until we can figure out how to effect meaningful change. I’ve collected some articles here about America’s dysfunctional relationship with guns, most of which I’ve shared before. Change is possible — there are good reasons to control the ownership of guns and control has a high likelihood of success — but how will our country find the political will to make it happen?

An armed society is not a free society:

Arendt offers two points that are salient to our thinking about guns: for one, they insert a hierarchy of some kind, but fundamental nonetheless, and thereby undermine equality. But furthermore, guns pose a monumental challenge to freedom, and particular, the liberty that is the hallmark of any democracy worthy of the name — that is, freedom of speech. Guns do communicate, after all, but in a way that is contrary to free speech aspirations: for, guns chasten speech.

This becomes clear if only you pry a little more deeply into the N.R.A.’s logic behind an armed society. An armed society is polite, by their thinking, precisely because guns would compel everyone to tamp down eccentric behavior, and refrain from actions that might seem threatening. The suggestion is that guns liberally interspersed throughout society would cause us all to walk gingerly — not make any sudden, unexpected moves — and watch what we say, how we act, whom we might offend.

We’re sacrificing America’s children to “our great god Gun”:

Read again those lines, with recent images seared into our brains — “besmeared with blood” and “parents’ tears.” They give the real meaning of what happened at Sandy Hook Elementary School Friday morning. That horror cannot be blamed just on one unhinged person. It was the sacrifice we as a culture made, and continually make, to our demonic god. We guarantee that crazed man after crazed man will have a flood of killing power readily supplied him. We have to make that offering, out of devotion to our Moloch, our god. The gun is our Moloch. We sacrifice children to him daily — sometimes, as at Sandy Hook, by directly throwing them into the fire-hose of bullets from our protected private killing machines, sometimes by blighting our children’s lives by the death of a parent, a schoolmate, a teacher, a protector. Sometimes this is done by mass killings (eight this year), sometimes by private offerings to the god (thousands this year).

The gun is not a mere tool, a bit of technology, a political issue, a point of debate. It is an object of reverence. Devotion to it precludes interruption with the sacrifices it entails. Like most gods, it does what it will, and cannot be questioned. Its acolytes think it is capable only of good things. It guarantees life and safety and freedom. It even guarantees law. Law grows from it. Then how can law question it?

Roger Ebert on the media’s coverage of mass shootings:

Let me tell you a story. The day after Columbine, I was interviewed for the Tom Brokaw news program. The reporter had been assigned a theory and was seeking sound bites to support it. “Wouldn’t you say,” she asked, “that killings like this are influenced by violent movies?” No, I said, I wouldn’t say that. “But what about ‘Basketball Diaries’?” she asked. “Doesn’t that have a scene of a boy walking into a school with a machine gun?” The obscure 1995 Leonardo Di Caprio movie did indeed have a brief fantasy scene of that nature, I said, but the movie failed at the box office (it grossed only $2.5 million), and it’s unlikely the Columbine killers saw it.

The reporter looked disappointed, so I offered her my theory. “Events like this,” I said, “if they are influenced by anything, are influenced by news programs like your own. When an unbalanced kid walks into a school and starts shooting, it becomes a major media event. Cable news drops ordinary programming and goes around the clock with it. The story is assigned a logo and a theme song; these two kids were packaged as the Trench Coat Mafia. The message is clear to other disturbed kids around the country: If I shoot up my school, I can be famous. The TV will talk about nothing else but me. Experts will try to figure out what I was thinking. The kids and teachers at school will see they shouldn’t have messed with me. I’ll go out in a blaze of glory.”

In short, I said, events like Columbine are influenced far less by violent movies than by CNN, the NBC Nightly News and all the other news media, who glorify the killers in the guise of “explaining” them. I commended the policy at the Sun-Times, where our editor said the paper would no longer feature school killings on Page 1. The reporter thanked me and turned off the camera. Of course the interview was never used. They found plenty of talking heads to condemn violent movies, and everybody was happy.

Jill Lepore on the United States of Guns:

There are nearly three hundred million privately owned firearms in the United States: a hundred and six million handguns, a hundred and five million rifles, and eighty-three million shotguns. That works out to about one gun for every American. The gun that T. J. Lane brought to Chardon High School belonged to his uncle, who had bought it in 2010, at a gun shop. Both of Lane’s parents had been arrested on charges of domestic violence over the years. Lane found the gun in his grandfather’s barn.

The United States is the country with the highest rate of civilian gun ownership in the world. (The second highest is Yemen, where the rate is nevertheless only half that of the U.S.) No civilian population is more powerfully armed. Most Americans do not, however, own guns, because three-quarters of people with guns own two or more. According to the General Social Survey, conducted by the National Policy Opinion Center at the University of Chicago, the prevalence of gun ownership has declined steadily in the past few decades. In 1973, there were guns in roughly one in two households in the United States; in 2010, one in three. In 1980, nearly one in three Americans owned a gun; in 2010, that figure had dropped to one in five.

A Land Without Guns: How Japan Has Virtually Eliminated Shooting Deaths:

The only guns that Japanese citizens can legally buy and use are shotguns and air rifles, and it’s not easy to do. The process is detailed in David Kopel’s landmark study on Japanese gun control, published in the 1993 Asia Pacific Law Review, still cited as current. (Kopel, no left-wing loony, is a member of the National Rifle Association and once wrote in National Review that looser gun control laws could have stopped Adolf Hitler.)

To get a gun in Japan, first, you have to attend an all-day class and pass a written test, which are held only once per month. You also must take and pass a shooting range class. Then, head over to a hospital for a mental test and drug test (Japan is unusual in that potential gun owners must affirmatively prove their mental fitness), which you’ll file with the police. Finally, pass a rigorous background check for any criminal record or association with criminal or extremist groups, and you will be the proud new owner of your shotgun or air rifle. Just don’t forget to provide police with documentation on the specific location of the gun in your home, as well as the ammo, both of which must be locked and stored separately. And remember to have the police inspect the gun once per year and to re-take the class and exam every three years.

Australia’s gun laws stopped mass shootings and reduced homicides, study finds:

From 1979 to 1996, the average annual rate of total non-firearm suicide and homicide deaths was rising at 2.1% per year. Since then, the average annual rate of total non-firearm suicide and homicide deaths has been declining by 1.4%, with the researchers concluding there was no evidence of murderers moving to other methods, and that the same was true for suicide.

The average decline in total firearm deaths accelerated significantly, from a 3% decline annually before the reforms to a 5% decline afterwards, the study found.

In the 18 years to 1996, Australia experienced 13 fatal mass shootings in which 104 victims were killed and at least another 52 were wounded. There have been no fatal mass shootings since that time, with the study defining a mass shooting as having at least five victims.

From The Onion, ‘No Way To Prevent This,’ Says Only Nation Where This Regularly Happens:

At press time, residents of the only economically advanced nation in the world where roughly two mass shootings have occurred every month for the past eight years were referring to themselves and their situation as “helpless.”

But America is not Australia or Japan. Dan Hodges said on Twitter a few years ago:

In retrospect Sandy Hook marked the end of the US gun control debate. Once America decided killing children was bearable, it was over.

This can’t be the last word on guns in America. We have to do better than this for our children and everyone else whose lives are torn apart by guns. But right now, we are failing them miserably, and Hodges’ words ring with the awful truth that all those lives and our diminished freedom & equality are somehow worth it to the United States as a society.

There's an impression among some people that "deep learning" has brought computer algorithms to the point where there's nothing left to do but to work out the details of further applications. This reminds me of what has been described as Ludwig Wittgenstein's belief in the early 1920s that the development of formal logic and the "picture theory" of meaning in his Tractatus Logico-Philosophicus reduced the elucidation (or dissolution) of all philosophical questions to a sort of clerical procedure.

Several recent articles, in different ways, call into question this modern view that Deep Learning (i.e. complex networks of linear algebra with interspersed point nonlinearities, whose millions or billions of parameters are automatically learned from digital examples) is a philosopher's stone whose application solves all algorithmic problems. Two among many others: Gary Marcus, "Deep Learning: A Critical Appraisal", arXiv.org 1/2/2018; Michael Jordan, "Artificial Intelligence — The Revolution Hasn’t Happened Yet", Medium 4/19/2018.

And two upcoming talks describe some of the remaining problems in speech and language technology.

This afternoon here at Penn, Michael Picheny's title is "Speech Recognition: What's Left?", at 1:30 in Levine 307:

Recent speech recognition advances on the SWITCHBOARD corpus suggest that because of recent advances in Deep Learning, we now achieve Word Error Rates comparable to human listeners. Does this mean the speech recognition problem is solved and the community can move on to a different set of problems? In this talk, we examine speech recognition issues that still plague the community and compare and contrast them to what is known about human perception. We specifically highlight issues in accented speech, noisy/reverberant speech, speaking style, rapid adaptation to new domains, and multilingual speech recognition. We try to demonstrate that compared to human perception, there is still much room for improvement, so significant work in speech recognition research is still required from the community.

For one simple example of an area where further R&D is needed, see our earlier discussion of the First DIHARD Speech Diarization Challenge, where on some input categories the winning system performed quite badly indeed, with frame-wise error rates in the 50% range even when the speech-segment regions were specified:

(See here for a description of the 10 eval-set data sources.)

And on Wednesday, Richard Sproat will be talking in room 7102 at the CUNY Graduate Center on the topic "Neural models of text normalization for speech applications":

Speech applications such as text-to-speech (TTS) or automatic speech recognition (ASR), must not only know how to read ordinary words, but must also know how to read numbers, abbreviations, measure expressions, times, dates, and a whole range of other constructions that one frequently finds in written texts. The problem of dealing with such material is called text normalization. The traditional approach to this problem, and the one currently used in Google’s deployed TTS and ASR systems, involves large hand-constructed grammars, which are costly to develop and tricky to maintain. It would be nice if one could simply train a system from text paired with its verbalization. I will present our work on applying neural sequence-to-sequence RNN models to the problem of text normalization. Given sufficient training data, such models can achieve very high accuracy, but also tend to produce the occasional error — reading “kB” as “hectare”, misreading a long number such as “3,281” — that would be problematic in a real application. The most powerful method we have found to correct such errors is to use finite-state over-generating covering grammars at decoding time to guide the RNN away from “silly” readings: Such covering grammars can be learned from a very small amount of annotated data. The resulting system is thus a hybrid system, rather than a purely neural one, a purely neural approach being apparently impossible at present.

This reminds me of something that happened in the mid-1980s, during one of the earlier waves of pseudo-neural over-enthusiasm. I was then at Bell Labs, and had presented at some conference or another a paper on our approach to translating text into instructions for the synthesizer. What we used then was a hybrid of a hand-coded text normalization system, a large pronouncing dictionary, dictionary extensions via analogical processes like inflection and rhyming, and hand-coded grapheme-to-phoneme transduction (though no neural nets at that time). Richard Sproat might well have been a co-author. Afterwards, someone came up to me and asked why we were bothering with all that hand-coding, since neural nets could learn better solutions to all such problems with no human intervention, citing Terry Sejnowski's work on learning g2p rules using an early NN architecture (Sejnowski & Rosenberg, "NETtalk: a parallel network that learns to read aloud", JHU EECS Technical Report 1986).

I say that they "asked me", but "berated me" was more like it. My interlocutor's level of fervent conviction was like someone arguing about scientology, socialism, salvation, or grammatical theories  — it had been mathematically proved, after all, that a non-linear perceptron with one hidden layer could asymptotically learn any finite computable function, more or less, so to solve a problem in any other way was clearly a sort of moral failure. Against this level of conviction, it was no help at all to point out that NETtalk's performance was actually not very good.

Not long thereafter, I moderated a 1986 ACL "Forum on Connectionism" (which is what pseudo-neural computation used to be called), and I think that my "Moderator Statement" (pdf here) has actually stood up pretty well over the intervening 32 years.

“Apps of a Feather” — a joint statement from the developers of several top third-party Twitter clients:

After August 16th, 2018, “streaming services” at Twitter will be removed. This means two things for third-party apps:

1. Push notifications will no longer arrive
2. Timelines won’t refresh automatically

If you use an app like Talon, Tweetbot, Tweetings, or Twitterrific, there is no way for its developer to fix these issues.

We are incredibly eager to update our apps. However, despite many requests for clarification and guidance, Twitter has not provided a way for us to recreate the lost functionality. We’ve been waiting for more than a year and have had one reprieve.

This antipathy to third-party clients is especially confounding considering that Twitter recently dropped support for their own native Mac client. As far as I’m aware, once this comes to pass next month, there will be no way to receive notifications of Twitter DMs on a Mac. None. (Twitter’s website doesn’t even support Safari’s desktop notification feature.) That’s just wacky.

Twitter management obviously wants to steer people to their first-party mobile app and desktop website. I get that. But they already have that: the overwhelming number of Twitter users use exactly those products to access the service. What Twitter management seems to be missing is that many of its most influential users — including yours truly, yes — have been on the platform a long time and have a high tendency to be among those who not just use, but depend upon third-party clients.

To me this is like finding out you’re now required to access email entirely through a web browser. Sure, lots of people already do it that way and either prefer it or think it’s eh, just fine, who cares — but a lot of others hate it and find it completely disruptive to longstanding workflows.

Twitter isn’t explicitly saying that they’re shutting down third-party clients, but I don’t know that it’s feasible for them to exist if they don’t have access to these APIs. It’s like breaking up with someone by being a jerk to them rather than telling them you’re breaking up.

I urge Twitter to reconsider this decision. Third-party clients account for a relatively small part of the Twitter ecosystem, but it’s an important one. Twitter may not care about a native Mac client, but the users of these apps, and the developers who make them, certainly do.

Fascinating, wide-ranging podcast interview (with an excellent transcript) between Vox’s Ezra Klein and Facebook CEO Mark Zuckerberg:

Klein: One of the things that has been coming up a lot in the conversation is whether the business model of monetizing user attention is what is letting in a lot of these problems. Tim Cook, the CEO of Apple, gave an interview the other day and he was asked what he would do if he was in your shoes. He said, “I wouldn’t be in this situation,” and argued that Apple sells products to users, it doesn’t sell users to advertisers, and so it’s a sounder business model that doesn’t open itself to these problems.

Do you think part of the problem here is the business model where attention ends up dominating above all else, and so anything that can engage has powerful value within the ecosystem?

Zuckerberg: You know, I find that argument, that if you’re not paying that somehow we can’t care about you, to be extremely glib and not at all aligned with the truth. The reality here is that if you want to build a service that helps connect everyone in the world, then there are a lot of people who can’t afford to pay. And therefore, as with a lot of media, having an advertising-supported model is the only rational model that can support building this service to reach people. […]

But if you want to build a service which is not just serving rich people, then you need to have something that people can afford. I thought Jeff Bezos had an excellent saying on this in one of his Kindle launches a number of years back. He said, “There are companies that work hard to charge you more, and there are companies that work hard to charge you less.” And at Facebook, we are squarely in the camp of the companies that work hard to charge you less and provide a free service that everyone can use.

I don’t think at all that that means that we don’t care about people. To the contrary, I think it’s important that we don’t all get Stockholm syndrome and let the companies that work hard to charge you more convince you that they actually care more about you. Because that sounds ridiculous to me.

There is certainly something to Zuckerberg’s argument here, but the speciousness of the way he formulates it is that a company working hard to charge you more money is undeniably incentivized to care more about you, if you can afford their product or service. I think it’s undeniable that Apple cares more about the hundreds of millions of people who buy its products than Facebook cares about any of its billions of users.

The more apt Tim Cook quote that applies here is this long-standing internet adage: “When an online service is free, you’re not the customer. You’re the product.” Amazon is undeniably focused on low prices. But Facebook doesn’t charge low prices — they charge high prices. To their customers: advertisers. And a cursory look at their financials indicates they’ve been working hard to raise those prices.

The linguistic trick Zuckerberg pulls here is that nowhere in the entire interview does he mention the words user or customer. He only says you (in the plural sense) and people. That’s a dodge, because unlike Apple — and Amazon — Facebook’s users are not its customers — and most of the controversies they are dealing with today all stem from the fact that they favored their customers (advertisers willing to pay ever-higher sums for ever-more-invasively-targeted ads) at the expense of their users.