Canada deputy high commissioner summoned after Sikh homeland slogans reportedly raised at an event addressed by PM.

Reading, while not technically medicine, is a fundamentally wholesome activity. It can prevent cognitive decline, improve sleep, and lower blood pressure. In one study, book readers outlived their nonreading peers by nearly two years. People have intuitively understood reading’s benefits for thousands of years: The earliest known library, in ancient Egypt, bore an inscription that read The house of healing for the soul.

But the ancients read differently than we do today. Until approximately the tenth century, when the practice of silent reading expanded thanks to the invention of punctuation, reading was synonymous with reading aloud. Silent reading was terribly strange, and, frankly, missed the point of sharing words to entertain, educate, and bond. Even in the 20th century, before radio and TV and smartphones and streaming entered American living rooms, couples once approached the evening hours by reading aloud to each other.

But what those earlier readers didn’t yet know was that all of that verbal reading offered additional benefits: It can boost the reader’s mood and ability to recall. It can lower parents’ stress and increase their warmth and sensitivity toward their children. To reap the full benefits of reading, we should be doing it out loud, all the time, with everyone we know.

Reading aloud is a distinctive cognitive process, more complex than simply reading silently, speaking, or listening. Noah Forrin, who researched memory and reading at the University of Waterloo, in Canada, told me that it involves several operations—motor control, hearing , and self-reference (the fact that you said it)—all of which activate the hippocampus, a brain region associated with episodic memory. Compared with reading silently, the hippocampus is more active while reading aloud, which might help explain why the latter is such an effective memory tool. In a small 2012 study, students who studied a word list remembered 90 percent of the words they’d read aloud immediately afterward, compared with 71 percent of those they’d read silently. (One week later, participants remembered 59 percent of the spoken words and 48 percent of the words read silently.)

So although you might enjoy an audiobook narrated by Meryl Streep, you would remember it better if you read parts of it out loud—especially if you did so in small chunks, just a short passage at a time, Forrin said. The same goes for a few lines of a presentation that you really want to nail. Those memory benefits hold true whether or not anyone is around to hear your performance.

Verbal reading without an audience is, in fact, surprisingly common. While studying how modern British people read aloud, Sam Duncan, a professor of adult literacies at University College London, found that they read aloud—and alone—for a variety of reasons. One woman recited Welsh poetry to remember her mother, with whom she spoke Welsh as a girl. One young man read the Quran out loud before work to better understand its meaning. Repeating words aloud isn’t just key to memorization, Duncan told me—it can be key to identity formation too.

[From the August 1904 issue: On reading aloud]

Plenty of solitary vocal reading no doubt consists of deciphering recipes and proofreading work emails, but if you want to reap the full perks, the best selections are poetry and literature. These genres provide access to facets of human experience that can be otherwise unreachable, which helps us process our own emotions and memories, says Philip Davis, an emeritus professor of literature and psychology at the University of Liverpool. Poetry, for example, can induce peak emotional responses, a strong reaction that might include goose bumps or chills. It can help you locate an emotion within yourself, which is important to health as a form of emotional processing.

Poetry also contains complex, unexpected elements, like when Shakespeare uses god as a verb in Coriolanus: “This last old man … godded me.” In an fMRI study that Davis co-authored in 2015, such literary surprise was shown to be stimulating to the brain. Davis told me that literature, with its “mixture of memory and imagination,” can cause us to recall our most complex experiences and derive meaning from them. A poem or story read aloud is particularly enthralling, he said, because it becomes a live presence in the room, with a more direct and penetrative quality, akin to live music. Davis likens the role of literature and live reading to a spark or renewal, “a bringing of things back to life.”

Discussing the literature that you read aloud can be particularly valuable. Davis told me doing so helps penetrate rigid thinking and can dislodge dysfunctional thought patterns. A qualitative study he co-authored in 2017 found that, for those who have chronic pain and the depression that tends to come with it, such discussion expands emotional vocabulary—a key tenet of psychological well-being—perhaps even more so than cognitive behavioral therapy. (The allure of an audience has one notable exception: If you’re anxious, reading aloud can actually reduce memory and comprehension. To understand this effect, one need only harken back to fifth grade when it was your turn to read a paragraph on Mesopotamia in class.)

[Read: How to keep your book club from becoming a wine club]

The health benefits of reading aloud are so profound that some doctors in England now refer their chronic-pain patients to read-aloud groups. Helen Cook, a 45-year-old former teacher in England, joined one of these groups in 2013. Cook had a pelvic tumor that had sent anguish ricocheting through her hip and back for a decade, and medication never seemed to help. Before she joined the reading group, Cook had trouble sleeping, lost her job, and “had completely lost myself,” she told me. Then, she and nine other adults began working their way through some 300 pages of Hard Times, by Charles Dickens.

Cook told me she recognized her experience in the characters’ travails, and within months, she “rediscovered a love for life,” even returning to college for a master’s degree in literature. She’s not the only one who found relief: In Davis’s 2017 study, everyone who read aloud in a group felt emotionally better and reported less pain for two days afterward.

Hearing words read aloud to you also has unique advantages, especially for kids. Storytelling has been shown to increase hospitalized children’s levels of oxytocin while decreasing cortisol and pain. Julie Hunter, who for more than 20 years has taught preschool kids (including my daughter), told me that interactive reading increases young children’s comprehension, builds trust, and enhances social-emotional skills. A recent study by researchers at the Brookings Institution found that children smiled and laughed more when being read to by a parent than when listening to an automatically narrated book alone.

[Read: An ode to being read to]

Anecdotal evidence suggests that adults, too, can benefit from such listening. For 25 years, Hedrick and Susan Smith, ages 90 and 84, respectively, have read more than 170 books aloud. They started by reading in the car, to pass the time, but it was so much fun that they started reading every night before they turned out the light, Hedrick told me. Together, they tried to comprehend One Hundred Years of Solitude, narrated Angela’s Ashes in four different Irish accents, and deduced clues in John le Carré thrillers. They felt more connected, and went to sleep in brighter moods. If they liked the book, they couldn’t wait for the other to read the next chapter aloud—even, and perhaps especially, when the sound of the other’s voice sent them off to sleep.

A train engulfed in flames rolled through a city in Ontario, Canada as passersby filmed it. No injuries were reported.

This is an edition of The Atlantic Daily, a newsletter that guides you through the biggest stories of the day, helps you discover new ideas, and recommends the best in culture. Sign up for it here.

Welcome back to The Daily’s Sunday culture edition, in which one Atlantic writer or editor reveals what’s keeping them entertained. Today’s special guest is Valerie Trapp, an assistant editor who has written about the adult stuffed-animal revival, a fun way to pick up a new language, and the long tradition of villain comedy.

Valerie is a “self-appointed emissary” for Crazy, Stupid, Love, which she calls “the perfect rom-com.” She loves listening to Bad Bunny’s “unfailing bangers,” will watch anything Issa Rae does, and was left in a brief stupor after reading The Order of Time by the physicist Carlo Rovelli.

First, here are three Sunday reads from The Atlantic:

Gavin Newsom can’t help himself. Welcome to pricing hell. “Nostalgia for a dating experience they’ve never had”

The Culture Survey: Valerie Trapp

The upcoming event I’m most looking forward to: I’m still riding a wave of postconcert bliss from the Bad Bunny tour, which left me wanting little. But if I could, I’d love to see the Shakira, Maggie Rogers, and Jazmine Sullivan tours, and Steve Carell and William Jackson Harper in the Uncle Vanya production on Broadway.

Something I recently revisited: I’ve been rereading the civil-rights lawyer Valarie Kaur’s memoir See No Stranger: A Memoir and Manifesto of Revolutionary Love. It’s an absolutely gorgeous and lucid guide on how to stretch our heart a little past what we think is possible. Kaur defines the act of wonder as looking at the world—trees, stars, people you do and don’t like—and thinking, “You are a part of me I do not yet know.” I return to such phrases when I need a way forward.

Best novel I’ve recently read, and the best work of nonfiction: The novel The Vulnerables, by Sigrid Nunez, entranced me with a voice I’d follow down any train of thought. And the physicist Carlo Rovelli’s The Order of Time left me walking around in a mild stupor for about 20 minutes, seeing buildings as events instead of as objects. Did I quickly forget all the physics Rovelli tried to teach me? I’d barely grasped it in the first place. But his poetic musings on how humans experience time and mortality have stayed with me. [Related: A new way to think about thinking]

Authors I will read anything by: Jia Tolentino, Maggie Nelson, Andrew Sean Greer, Joy Harjo, Michael Pollan.

My favorite blockbuster and favorite art movie: Maybe not a blockbuster, but I’ll mention it anyway, because I am its self-appointed emissary: Crazy, Stupid, Love is the perfect rom-com. It’s a Shakespearean comedy of errors with jokes about the Gap and many perfect uses of the word cuckold. Could we ask for more? As for an art film, I love Pedro Almodóvar’s Parallel Mothers—Penélope Cruz is brilliant in it (and in pretty much everything she does).

An actor I would watch in anything: In college, I was fascinated by Margot Robbie’s “animal work” method-acting process, which involves studying and embodying different animals to shape the physicality of her roles. She prepared for I, Tonya by observing bulldogs and wild horses; for Babylon, she studied octopi and honey badgers! I had a philosophy professor in college who once made us do a similar exercise as homework. I ended up embodying a crow, and by this I mean I made a gigantic fool of myself by squawking in front of passersby. So props to Margot—I’m happy to sit that exercise out and watch her do it instead.

A quiet song that I love, and a loud song that I love: A quiet song: “Rodeo Clown,” by Dijon. I’ll play the entirety of Dijon’s discography when I feel even a bit moody, and this song is the pinnacle of moodiness. It’s perfect and a little deranged, all soul and catharsis. “You’re missin’ out on some good, good lovin’!” Dijon wails, screeching and theatrical, shortly before an interlude of quiet sobs.

A loud song: “Safaera,” by Bad Bunny, Jowell & Randy, and Ñengo Flow. Bad Bunny makes unfailing bangers that switch up and crescendo, taking you on a complete and adequately tiring perreo journey. “Todo Tiene Su Hora,” by Juan Luis Guerra, also can get me both dancing and crying happy tears of wonder at the magic of the world. [Related: Bad Bunny overthrows the Grammys.]

A musical artist who means a lot to me: Beyoncé. She’s ecstatic and lavish in her artistry. I think sometimes about a moment in her documentary Life Is but a Dream in which she emphatically tells a crowd, “I’m gonna give you everything I have. I promise!” I find that kind of exuberant generosity very moving.

A favorite story I’ve read in The Atlantic: Recently, Sarah Zhang’s article about the life-changing effects of a cystic-fibrosis breakthrough and Ross Andersen’s story about our hypothetical contact with whale civilizations left me in absolute awe.

The last entertainment thing that made me cry: I might not be the best gauge for this question, because I cry easily and for most movies—including once during a viewing of Justin Bieber’s 2013 concert film. But recently: the song “2012,” by Saba. It feels like time travel and sounds like nostalgia. It was the sweeping post-chorus, which speaks to simpler days, that got me: “I had everything I needed, everything / ’Cause I had everyone I needed.”

An online creator that I’m a fan of: I’m a devoted reader of Heather Havrilesky’s Ask Polly Substack, which is consistently hilarious, comforting, and sharp.

A good recommendation I recently received: Young Miko’s new album, Att.—it’s a 46-minute-long party. I saved pretty much every track and especially loved “ID” and “Fuck TMZ.”

The last thing that made me snort with laughter: I started rewatching Insecure this year while doing my taxes. Dare I say, I almost had a nice time on TurboTax. The show’s pilot remains brilliant. The “Broken Pussy” rap remains hilarious. I will watch anything Issa Rae does. [Related: How Issa Rae built the world of Insecure]

A poem, or line of poetry, that I return to: I mutter lines from “The Story Wheel,” by Joy Harjo, like affirmations. Whenever I feel myself slipping into self-deprecation or pride, I recall: “None of us is above the other / In this story of forever. / Though we follow that red road home, / one behind another.”

The Week Ahead

Challengers, a film directed by Luca Guadagnino about a former tennis star turned coach, played by Zendaya, who is enmeshed in a love triangle with two pro players (in theaters Friday) The Jinx: Part Two, the second installment of the infamous true-crime docuseries, in which the real-estate heir Robert Durst seemingly confessed to murder (premieres today on HBO and Max) Funny Story, a book by Emily Henry about a woman whose life is upended when her fiancé leaves her for his childhood best friend (out Tuesday)

Essay

Getty

The Most Hated Sound on Television

By Jacob Stern

Viewers scorned the laugh track—prerecorded and live chortles alike—first for its deceptiveness and then for its condescension. They came to see it as artificial, cheesy, even insulting: You think we need you to tell us when to laugh? Larry Gelbart said he “always thought it cheapened” M*A*S*H. Larry David reportedly didn’t want it on Seinfeld but lost out to studio execs who did. The actor David Niven once called it “the single greatest affront to public intelligence I know of.” In 1999, Time judged the laugh track to be “one of the hundred worst ideas of the twentieth century.” And yet, it persisted. Until the early 2000s, nearly every TV comedy relied on one. Friends, Two and a Half Men, Everybody Loves Raymond, Drake & Josh—they all had laugh tracks.

Now the laugh track is as close to death as it’s ever been.

Read the full article.

More in Culture

Taylor Swift is having quality-control issues. Eight cookbooks worth reading cover to cover The uncomfortable truth about child abuse in Hollywood The illogical relationship Americans have with animals Something weird is happening with Caesar salads. Short story: “The Vale of Cashmere” Is this the end for Bluey? Prestige TV’s new wave of difficult men Ken will never die.

Catch Up on The Atlantic

The Trump trial’s extraordinary opening Finding justice in Palestine Why did U.S. planes defend Israel but not Ukraine?

Photo Album

Theo Dagnaud, a member of a fire crew, scans the horizon during Canada’s recent summer of gigantic forest fires. (Charles-Frederick Ouellet)

Check out the winning entries of this year’s World Press Photo Contest, including images of the aftermath of an earthquake in Turkey, Canada’s scorching wildfire summer, and war in Gaza.

Explore all of our newsletters.

When you buy a book using a link in this newsletter, we receive a commission. Thank you for supporting The Atlantic.

Mass shootings didn’t start at Columbine High, but the mass-shooter era did. Eric Harris and Dylan Klebold’s audacious plan and misread motives multiplied the stakes and inspired wave after wave of emulation. How could we know we were witnessing an origin story?

The legend of Columbine is fiction. There are two versions of the attack: what actually happened on April 20, 1999, and the story we all accepted back then. The mythical version explained it all so cleanly. A pair of outcast loners dubbed the “Trench Coat Mafia” targeted the jocks to avenge years of bullying. Dwayne Fuselier, the supervisory special agent who led the FBI’s Columbine investigation, is fond of quoting H. L. Mencken in response to the mythmaking: “There is always a well-known solution to every human problem—neat, plausible, and wrong.”

The legend hinges on bullying, but the killers never mentioned it in the huge trove of journals, online posts, and videos they left to explain themselves. The myth was so insidious because it cast the ruthless killers as heroes of misfits everywhere. Fuselier warned how appealing that myth would sound to anyone who felt ostracized. Within a few years, the fledgling fandom would find one another on social media, where they have operated ever since.

Around the world, Eric and Dylan are idolized as champions of “the nobodies.” Eric hated the nobodies. He mocked them mercilessly on his website and in his journal. He wasn’t a loner or an outcast, and neither was Dylan. Eric and Dylan made clear in their writings that they were planning the attack for their own selfish motives—certainly not to help the kids they ridiculed at the bottom of the social food chain.

They were not in the Trench Coat Mafia. They were not Nazis or white supremacists, and they did not plan the attack for Hitler’s birthday. They did not target jocks, Christians, or Black people. They targeted no one specifically. They shot randomly and designed their bombs to kill indiscriminately. That’s where “they” ends: Their polar-opposite personalities drove opposite motives. Psychopaths are devoid of empathy; Eric was a sadistic psychopath who killed for his own aggrandizement and enjoyment. Dylan was suicidally depressed and self-loathing. Eric lured him into punishing the world for the pain it inflicted on him, instead of punishing himself. Columbine was a suicide plan, but on “Judgment Day,” as they called it, Dylan would show the world the “somebody” we’d never seen.

The Columbine killers have fans. Eric and Dylan’s adoring online following spreads across nearly every continent, and it’s growing across multiple platforms. In Russia, the government, which has been plagued by an explosion of both Columbine fandom and mass shootings, estimates that more than 70,000 members exist. They call themselves the TCC, for “True Crime Community,” and I’ve spent much of the past 15 years inside their online world. My book Columbine made me enemy No. 1 for portraying Eric and Dylan as ruthless murderers.

In 2016, a young fan tweeted: “hey @DaveCullen block me or else i shoot my school.” She’d been ranting for hours, posting pictures of school shooters, and tweets such as: “It’s also something a lot of people need, To die....I wish i was dead...I LIKE VIOLENCE...I want to be killed in front of an audience. … I think someone failed to abort me (:”

These teens are ensnared in an American tragedy that just keeps growing worse.

I’ve tried to leave this story so many times, but this diagram haunts me, ruthlessly expanding like an unstoppable spider web, devouring all the lives and futures in its path. It demands that we address the cause—25 years too late. That web is made up of 54 mass shootings that have killed nearly 300 people and wounded more than 500. And every gunman left evidence that they were inspired or influenced by the murderers at Columbine. The Columbine effect.

Eric and Dylan’s bombs failed. Yet the legend made them heroic to their progeny and gave birth to their fandom. By the tenth anniversary, a small band of “Columbiners” had formed online. They gravitated to the TCC, to Ted Bundy, to the younger Tsarnaev brother, ‎to Dylann Roof, and to others—but Eric and Dylan are the megastars. The groupies multiply, as fresh crops of teens join their ranks each season.

Most gunmen die in the act, so the 54 attacks itemized in the diagram are just the ones that we know of, and that were carried out. A 2015 Mother Jones investigation of Columbine copycats found more than two thwarted attacks for each one that succeeded. It identified 14 plotters targeting Columbine’s anniversary and 13 striving to top its body count. Surviving mass shooters have admitted that they were competing with one another.

All roads lead back to Columbine. The Virginia Tech shooter, Seung-Hui Cho, wrote in a school assignment that he wanted to “repeat Columbine” and that he idolized its “martyrs.” The Northern Illinois University killer marked a third generation, explicitly inspired by both Virginia Tech and Columbine. Sandy Hook was the fourth generation; Adam Lanza had studied all three. Six more school shooters later referenced Sandy Hook and Columbine. Five generations of fallout, all reenacting the original legend.

Most early Columbiners were just curious teenagers interested in the criminal mind or in analyzing Columbine. Many still are, and their analyses are often useful. Many are angry about being tarred with the group’s reputation, but they have been outnumbered by new arrivals unabashedly calling themselves fans. Many use the killers’ faces as avatars, extoll their virtues, and compose love poems, fan fiction, and gory memes about them. Sue Klebold said she was shocked by the volume of letters she received calling Dylan “heroic” and by the number of girls saying, “I wish I could have his baby.”

How little these groupies know about the murderers they obsess over is ironic. They keep repeating the misreporting that was debunked decades ago, convinced it’s true because it has metastasized into TCC dogma. The TCC twists the story to recast the murderers as victims; and the dead, wounded, and traumatized as villains. The groupies didn’t start these myths; we in the media bear that shame. But the groupies are now the carriers, spreading the legend of Dylan and Eric to remote reaches of the globe.

Seventy thousand is a tiny fraction of the adolescent population, but a magnet for a dangerous cohort of marginalized, disaffected, and hopeless teens—a major pool of aspiring shooters. Most TCC members outright say that they condone the Columbine murders, often in their profiles. They have turned Eric and Dylan into folk heroes, and they celebrate them as avenging angels. Adam Lanza obsessed over the Columbine killers and spent years immersed in these groups online. Then he murdered 20 little kids and six adults at Sandy Hook Elementary School.

Here’s the twist: Most of the TCC members I’ve engaged with describe themselves as awkward outcasts desperate to fit in. The TCC embraces them. The TCC feels cool—Eric and Dylan are super cool—and so they finally feel cool. I find it heartbreaking to hear them describe the pain they endure at school and the affinity they feel for “Dylan” and “Eric,” the fictional characters they’ve constructed. These kids are shocked when I tell them that other members of the TCC have told me the same—that they are putting on the same show, sure that all the others really mean it. Did Adam Lanza believe the posers? We’ll never know, but we can be certain that as you read this, a distraught, lonely kid somewhere is contemplating an attack—and the one community they trust is screaming, Do it!

Lots of kids fantasize about killing. Two days after Columbine, Salon ran “Misfits Who Don’t Kill,” in which three people came clean about their youthful fantasies of enacting mass murder. The phenomenon was widely reported that week. But none of those people did anything, because they knew how horribly wrong acting out the fantasy would be. Inside the TCC bubble, the constant message is that if your classmates are tormenting you, killing them is not just moral —it’s heroic and noble.

The TCC has a tell: Actual shootings unnerve them. Their posts grow quiet, respectful, and even mournful after some troubled young person heeds their call. I can gauge the change instantly, because the incessant harassment I get from them stops cold—for a week or two. Parkland was different: Six months went by before the taunts began trickling back in, and I haven’t gotten a death threat in the six years since. Why? I have no way to be certain about this, but my educated guess is that David Hogg, X González, and the rest of the March for Our Lives kids were suddenly cooler than the young shooters. And so much more powerful.

Eric and Dylan weren’t powerful—their plan failed. They’d planned Columbine as a bombing, the primary terrorist tactic. They thought they were launching a three-act drama: The cafeteria bombs would kill nearly 600 people instantly; what they called the “fun” part would be shooting up hundreds of survivors; and the massive car bombs set in the parking lot outside were to be the coup de grâce. Those timers were set to explode 45 minutes after the initial blast, wiping out countless more survivors and first responders, live on national TV. The Columbine killers’ performance was staged as the most apocalyptic made-for-TV horror film in American history. Eric complained in his journal that his “audience” would fail to understand. He got that right. He got everything else wrong.

Every element fizzled. All of the big bombs failed. Eric and Dylan went down to the cafeteria in a last desperate move to ignite the bombs with gunfire and a Molotov cocktail. Failed. Experts on psychopaths say they get bored after their initial kills, and Eric had likely lost interest. His gun’s recoil had broken his nose, so he spent that time in acute pain. The cops refused to kill them in the blaze of glory that they’d described as their final curtain. The smell of all the blood and already decomposing bodies was overpowering. Out of options, each shot himself in the head.

A more obscene and pathetic way to die is hard to imagine. Yet their fans have never confronted that ugly reality, because the opposite story took hold, making Eric and Dylan masterminds of the “worst school shooting in American history.”

The Columbine effect has gone global. It has inspired mass shootings in Finland, Sweden, Brazil, Mexico, Canada, France, Germany, the Netherlands, Ukraine, and Russia—as well as knife and axe attacks in places as remote as Siberia. In 2022, Russia designated the online “Columbine movement” a terrorist group. To comply with the ruling, my publisher required me to disavow the group in the Russian translation of Columbine. Mass murder inspired by those inept perpetrators is America’s most revolting cultural export.

I know when the TCC colonizes a new region, because I start getting a barrage of taunts in a different language. It’s a social contagion. Researchers have described school shootings as the American equivalent of suicide bombings—an ideology joined with a tactic. The phenomenon is escalating and self-perpetuating.

The Columbine groupies have no idea that they’re exporting a fraud. The media set this whole thing in motion 25 years ago. To untell a legend is a formidable task. It will be possible only when the media finally begin to convey how pathetic and gruesome the killers’ final moments were. The fans need to hear the ugly truth. Eric and Dylan viciously murdered innocent kids for their own selfish and petty agendas, and they died miserable failures.

This essay is adapted by the author from the new preface to a 25th-anniversary edition of Columbine.

Exactly a year after 6,600 bars of 99.99-percent pure gold were reported missing at Toronto Pearson International Airport, Peel Regional Police has announced multiple arrests in connection with the brazen heist seemingly ripped straight from the silver screen. Law enforcement claims two Air Canada employees were…

Read more...

Student activists at a university in Canada have ‘renamed’ dozens of campus buildings to places in Palestine.

A year after millions in gold bars and cash were stolen from a Toronto airport, police offer new details.

Budget proposes higher capital gains taxes and billions in spending on education, housing, jobs and mental health.

Recently, Bonaventure Dossou learned of an alarming tendency in a popular AI model. The program described Fon—a language spoken by Dossou’s mother and millions of others in Benin and neighboring countries—as “a fictional language.”

This result, which I replicated, is not unusual. Dossou is accustomed to the feeling that his culture is unseen by technology that so easily serves other people. He grew up with no Wikipedia pages in Fon, and no translation programs to help him communicate with his mother in French, in which he is more fluent. “When we have a technology that treats something as simple and fundamental as our name as an error, it robs us of our personhood,” Dossou told me.

The rise of the internet, alongside decades of American hegemony, made English into a common tongue for business, politics, science, and entertainment. More than half of all websites are in English, yet more than 80 percent of people in the world don’t speak the language. Even basic aspects of digital life—searching with Google, talking to Siri, relying on autocorrect, simply typing on a smartphone—have long been closed off to much of the world. And now the generative-AI boom, despite promises to bridge languages and cultures, may only further entrench the dominance of English in life on and off the web.

Scale is central to this technology. Compared with previous generations, today’s AI requires orders of magnitude more computing power and training data, all to create the humanlike language that has bedazzled so many users of ChatGPT and other programs. Much of the information that generative AI “learns” from is simply scraped from the open web. For that reason, the preponderance of English-language text online could mean that generative AI works best in English, cementing a cultural bias in a technology that has been marketed for its potential to “benefit humanity as a whole.” Some other languages are also well positioned for the generative-AI age, but only a handful: Nearly 90 percent of websites are written in just 10 languages (English, Russian, Spanish, German, French, Japanese, Turkish, Portuguese, Italian, and Persian).

Some 7,000 languages are spoken in the world. Google Translate supports 133 of them. Chatbots from OpenAI, Google, and Anthropic are still more constrained. “There’s a sharp cliff in performance,” Sara Hooker, a computer scientist and the head of Cohere for AI, a nonprofit research arm of the tech company Cohere, told me. “Most of the highest-performance [language] models serve eight to 10 languages. After that, there’s almost a vacuum.” As chatbots, translation devices, and voice assistants become a crucial way to navigate the web, that rising tide of generative AI could wash out thousands of Indigenous and low-resource languages such as Fon—languages that lack sufficient text with which to train AI models.  

“Many people ignore those languages, both from a linguistic standpoint and from a computational standpoint,” Ife Adebara, an AI researcher and a computational linguist at the University of British Columbia, told me. Younger generations will have less and less incentive to learn their forebears’ tongues. And this is not just a matter of replicating existing issues with the web: If generative AI indeed becomes the portal through which the internet is accessed, then billions of people may in fact be worse off than they are today.

Adebara and Dossou, who is now a computer scientist at Canada’s McGill University, work with Masakhane, a collective of researchers building AI tools for African languages. Masakhane, in turn, is part of a growing, global effort racing against the clock to create software for, and hopefully save, languages that are poorly represented on the web. In recent decades, “there has been enormous progress in modeling low-resource languages,” Alexandra Birch, a machine-translation researcher at the University of Edinburgh, told me.

In a promising development that speaks to generative AI’s capacity to surprise, computer scientists have discovered that some AI programs can pinpoint aspects of communication that transcend a specific language. Perhaps the technology could be used to make the web more aware of less common tongues. A program trained on languages for which a decent amount of data are available—English, French, or Russian, say—will then perform better in a lower-resourced language, such as Fon or Punjabi. “Every language is going to have something like a subject or a verb,” Antonios Anastasopoulos, a computer scientist at George Mason University, told me. “So even if these manifest themselves in very different ways, you can learn something from all of the other languages.” Birch likened this to how a child who grows up speaking English and German can move seamlessly between the two, even if they haven’t studied direct translations between the languages—not moving from word to word, but grasping something more fundamental about communication.

[Read: The end of foreign-language education]

But this discovery alone may not be enough to turn the tide. Building AI models for low-resource languages is painstaking and time-intensive. Cohere recently released a large language model that has state-of-the-art performance for 101 languages, of which more than half are low-resource. That leaves about 6,900 languages to go, and this effort alone required 3,000 people working across 119 countries. To create training data, researchers frequently work with native speakers who answer questions, transcribe recordings, or annotate existing text, which can be slow and expensive. Adebara spent years curating a 42-gigabyte training data set for 517 African languages, the largest and most comprehensive to date. Her data set is 0.4 percent of the size of the largest publicly available English training data set. OpenAI’s proprietary databases—the ones used to train products such as ChatGPT—are likely far larger.

Much of the limited text readily available in low-resource languages is of poor quality—itself badly translated—or limited use. For years, the main sources of text for many such low-resource languages in Africa were translations of the Bible or missionary websites, such as those from Jehovah’s Witnesses. And crucial examples for fine-tuning AI, which has to be intentionally created and curated—data used to make a chatbot helpful, human-sounding, not racist, and so on—are even rarer. Funding, computing resources, and language-specific expertise are frequently just as hard to come by. Language models can struggle to comprehend non-Latin scripts or, because of limited training examples, to properly separate words in low-resource-language sentences—not to mention those without a writing system.

The trouble is that, while developing tools for these languages is slow going, generative AI is rapidly overtaking the web. Synthetic content is flooding search engines and social media like a kind of gray goo, all in hopes of making a quick buck.

Most websites make money through advertisements and subscriptions, which rely on attracting clicks and attention. Already, an enormous portion of the web consists of content with limited literary or informational merit—an endless ocean of junk that exists only because it might be clicked on. What better way to expand one’s audience than to translate content into another language with whatever AI program comes up on a Google search?

[Read: Prepare for the textpocalypse]

Those translation programs, already of sometimes questionable accuracy, are especially bad with low-resourced languages. Sure enough, researchers published preliminary findings earlier this year that online content in such languages was more likely to have been (poorly) translated from another source, and that the original material was itself more likely to be geared toward maximizing clicks, compared with websites in English or other higher-resource languages. Training on large amounts of this flawed material will make products such as ChatGPT, Gemini, and Claude even worse for low-resource languages, akin to asking someone to prepare a fresh salad with nothing more than a pound of ground beef. “You are already training the model on incorrect data, and the model itself tends to produce even more incorrect data,” Mehak Dhaliwal, a computer scientist at UC Santa Barbara and one of the study’s authors, told me—potentially exposing speakers of low-resource languages to misinformation. And those outputs, spewed across the web and likely used to train future language models, could create a feedback loop of degrading performance for thousands of languages.

Imagine “you want to do a task, and you want a machine to do it for you,” David Adelani, a DeepMind research fellow at University College London, told me. “If you express this in your own language and the technology doesn’t understand, you will not be able to do this. A lot of things that simplify lives for people in economically rich countries, you will not be able to do.” All of the web’s existing linguistic barriers will rise: You won’t be able to use AI to tutor your child, draft work memos, summarize books, conduct research, manage a calendar, book a vacation, fill out tax forms, surf the web, and so on. Even when AI models are able to process low-resource languages, the programs require more memory and computational power to do so, and thus become significantly more expensive to run—meaning worse results at higher costs.

AI models might also be void of cultural nuance and context, no matter how grammatically adept they become. Such programs long translated “good morning” to a variation of “someone has died” in Yoruba, Adelani said, because the same Yoruba phrase can convey either meaning. Text translated from English has been used to generate training data for Indonesian, Vietnamese, and other languages spoken by hundreds of millions of people in Southeast Asia. As Holy Lovenia, a researcher at AI Singapore, the country’s program for AI research, told me, the resulting models know much more about hamburgers and Big Ben than local cuisines and landmarks.

It may already be too late to save some languages. As AI and the internet make English and other higher-resource languages more and more convenient for young people, Indigenous and less widely spoken tongues could vanish. If you are reading this, there is a good chance that much of your life is already lived online; that will become true for more people around the world as time goes on and technology spreads. For the machine to function, the user must speak its language.

By default, less common languages may simply seem irrelevant to AI, the web, and, in turn, everyday people—eventually leading to abandonment. “If nothing is done about this, it could take a couple of years before many languages go into extinction,” Adebara said. She is already witnessing languages she studied as an undergraduate dwindle in their usage. “When people see that their languages have no orthography, no books, no technology, it gives them the impression that their languages are not valuable.”

[Read: AI is exposing who really has power in Silicon Valley]

Her own work, including a language model that can read and write in hundreds of African languages, aims to change that. When she shows speakers of African languages her software, they tell her, “‘I saw my language in the technology you built; I wasn’t expecting to see it there,’” Adebara said. “‘I didn’t know that some technology would be able to understand some part of my language,’ and they feel really excited. That makes me also feel excited.”

Several experts told me that the path forward for AI and low-resource languages lies not only in technical innovation, but in just these sorts of conversations: not indiscriminately telling the world it needs ChatGPT, but asking native speakers what the technology can do for them. They might benefit from better voice recognition in a local dialect, or a program that can read and digitize non-Roman script, rather than the all-powerful chatbots being sold by tech titans. Rather than relying on Meta or OpenAI, Dossou told me, he hopes to build “a platform that is appropriate and proper to African languages and Africans, not trying to generalize as Big Tech does.” Such efforts could help give low-resource languages a presence on the internet where there was almost none before, for future generations to use and learn from.

Today, there is a Fon Wikipedia, although its 1,300 or so articles are about two-thousandths of the total on its English counterpart. Dossou has worked on AI software that does recognize names in African languages. He translated hundreds of proverbs between French and Fon manually, then created a survey for people to tell him common Fon sentences and phrases. The resulting French-Fon translator he built has helped him better communicate with his mother—and his mother’s feedback on those translations has helped improve the AI program. “I would have needed a machine-translation tool to be able to communicate with her,” he said. Now he is beginning to understand her without machine assistance. A person and their community, rather than the internet or a piece of software, should decide their native language—and Dossou is realizing that his is Fon, rather than French.