Group items matching

in title, tags, annotations or url

start with epistemic bubbles

That omission might be purposeful

But that omission can also be entirely inadvertent. Even if we’re not actively trying to avoid disagreement, our Facebook friends tend to share our views and interests

An ‘echo chamber’ is a social structure from which other relevant voices have been actively discredited. Where an epistemic bubble merely omits contrary views, an echo chamber brings its members to actively distrust outsiders.

an echo chamber is something like a cult. A cult isolates its members by actively alienating them from any outside sources. Those outside are actively labelled as malignant and untrustworthy.

In epistemic bubbles, other voices are not heard; in echo chambers, other voices are actively undermined.

The way to break an echo chamber is not to wave “the facts” in the faces of its members. It is to attack the echo chamber at its root and repair that broken trust.

Looking to others for corroboration is a basic method for checking whether one has reasoned well or badly

They have been in the limelight lately, most famously in Eli Pariser’s The Filter Bubble (2011) and Cass Sunstein’s #Republic: Divided Democracy in the Age of Social Media (2017).

The general gist: we get much of our news from Facebook feeds and similar sorts of social media. Our Facebook feed consists mostly of our friends and colleagues, the majority of whom share our own political and cultural views

various algorithms behind the scenes, such as those inside Google search, invisibly personalise our searches, making it more likely that we’ll see only what we want to see. These processes all impose filters on information.

Such filters aren’t necessarily bad. The world is overstuffed with information, and one can’t sort through it all by oneself: filters need to be outsourced.

That’s why we all depend on extended social networks to deliver us knowledge

any such informational network needs the right sort of broadness and variety to work

Each individual person in my network might be superbly reliable about her particular informational patch but, as an aggregate structure, my network lacks what Sanford Goldberg in his book Relying on Others (2010) calls ‘coverage-reliability’. It doesn’t deliver to me a sufficiently broad and representative coverage of all the relevant information.

Epistemic bubbles also threaten us with a second danger: excessive self-confidence.

An ‘epistemic bubble’ is an informational network from which relevant voices have been excluded by omission

Suppose that I believe that the Paleo diet is the greatest diet of all time. I assemble a Facebook group called ‘Great Health Facts!’ and fill it only with people who already believe that Paleo is the best diet. The fact that everybody in that group agrees with me about Paleo shouldn’t increase my confidence level one bit. They’re not mere copies – they actually might have reached their conclusions independently – but their agreement can be entirely explained by my method of selection.

Luckily, though, epistemic bubbles are easily shattered. We can pop an epistemic bubble simply by exposing its members to the information and arguments that they’ve missed.

echo chambers are a far more pernicious and robust phenomenon.

amieson and Cappella’s book is the first empirical study into how echo chambers function

echo chambers work by systematically alienating their members from all outside epistemic sources.

Their research centres on Rush Limbaugh, a wildly successful conservative firebrand in the United States, along with Fox News and related media

His constant attacks on the ‘mainstream media’ are attempts to discredit all other sources of knowledge. He systematically undermines the integrity of anybody who expresses any kind of contrary view.

outsiders are not simply mistaken – they are malicious, manipulative and actively working to destroy Limbaugh and his followers. The resulting worldview is one of deeply opposed force, an all-or-nothing war between good and evil

The result is a rather striking parallel to the techniques of emotional isolation typically practised in cult indoctrination

cult indoctrination involves new cult members being brought to distrust all non-cult members. This provides a social buffer against any attempts to extract the indoctrinated person from the cult.

The echo chamber doesn’t need any bad connectivity to function. Limbaugh’s followers have full access to outside sources of information

As Elijah Millgram argues in The Great Endarkenment (2015), modern knowledge depends on trusting long chains of experts. And no single person is in the position to check up on the reliability of every member of that chain

Their worldview can survive exposure to those outside voices because their belief system has prepared them for such intellectual onslaught.

exposure to contrary views could actually reinforce their views. Limbaugh might offer his followers a conspiracy theory: anybody who criticises him is doing it at the behest of a secret cabal of evil elites, which has already seized control of the mainstream media.

Perversely, exposure to outsiders with contrary views can thus increase echo-chamber members’ confidence in their insider sources, and hence their attachment to their worldview.

‘evidential pre-emption’. What’s happening is a kind of intellectual judo, in which the power and enthusiasm of contrary voices are turned against those contrary voices through a carefully rigged internal structure of belief.

One might be tempted to think that the solution is just more intellectual autonomy. Echo chambers arise because we trust others too much, so the solution is to start thinking for ourselves.

that kind of radical intellectual autonomy is a pipe dream. If the philosophical study of knowledge has taught us anything in the past half-century, it is that we are irredeemably dependent on each other in almost every domain of knowledge

Limbaugh’s followers regularly read – but do not accept – mainstream and liberal news sources. They are isolated, not by selective exposure, but by changes in who they accept as authorities, experts and trusted sources.

we depend on a vastly complicated social structure of trust. We must trust each other, but, as the philosopher Annette Baier says, that trust makes us vulnerable. Echo chambers operate as a kind of social parasite on that vulnerability, taking advantage of our epistemic condition and social dependency.

I am quite confident that there are plenty of echo chambers on the political Left. More importantly, nothing about echo chambers restricts them to the arena of politics

The world of anti-vaccination is clearly an echo chamber, and it is one that crosses political lines. I’ve also encountered echo chambers on topics as broad as diet (Paleo!), exercise technique (CrossFit!), breastfeeding, some academic intellectual traditions, and many, many more

Here’s a basic check: does a community’s belief system actively undermine the trustworthiness of any outsiders who don’t subscribe to its central dogmas? Then it’s probably an echo chamber.

much of the recent analysis has lumped epistemic bubbles together with echo chambers into a single, unified phenomenon. But it is absolutely crucial to distinguish between the two.

Epistemic bubbles are rather ramshackle; they go up easily, and they collapse easily

Echo chambers are far more pernicious and far more robust. They can start to seem almost like living things. Their belief systems provide structural integrity, resilience and active responses to outside attacks

the two phenomena can also exist independently. And of the events we’re most worried about, it’s the echo-chamber effects that are really causing most of the trouble.

new data does, in fact, seem to show that people on Facebook actually do see posts from the other side, or that people often visit websites with opposite political affiliation.

their basis for evaluation – their background beliefs about whom to trust – are radically different. They are not irrational, but systematically misinformed about where to place their trust.

Many people have claimed that we have entered an era of ‘post-truth’.

Not only do some political figures seem to speak with a blatant disregard for the facts, but their supporters seem utterly unswayed by evidence. It seems, to some, that truth no longer matters.

This is an explanation in terms of total irrationality. To accept it, you must believe that a great number of people have lost all interest in evidence or investigation, and have fallen away from the ways of reason.

echo chambers offers a less damning and far more modest explanation. The apparent ‘post-truth’ attitude can be explained as the result of the manipulations of trust wrought by echo chambers.

We don’t have to attribute a complete disinterest in facts, evidence or reason to explain the post-truth attitude. We simply have to attribute to certain communities a vastly divergent set of trusted authorities.

An echo chamber doesn’t destroy their members’ interest in the truth; it merely manipulates whom they trust and changes whom they accept as trustworthy sources and institutions.

in many ways, echo-chamber members are following reasonable and rational procedures of enquiry. They’re engaging in critical reasoning. They’re questioning, they’re evaluating sources for themselves, they’re assessing different pathways to information. They are critically examining those who claim expertise and trustworthiness, using what they already know about the world

none of this weighs against the existence of echo chambers. We should not dismiss the threat of echo chambers based only on evidence about connectivity and exposure.

Notice how different what’s going on here is from, say, Orwellian doublespeak, a deliberately ambiguous, euphemism-filled language designed to hide the intent of the speaker.

echo chambers don’t trade in vague, ambiguous pseudo-speech. We should expect that echo chambers would deliver crisp, clear, unambiguous claims about who is trustworthy and who is not

clearly articulated conspiracy theories, and crisply worded accusations of an outside world rife with untrustworthiness and corruption.

Once an echo chamber starts to grip a person, its mechanisms will reinforce themselves.

In an epistemically healthy life, the variety of our informational sources will put an upper limit to how much we’re willing to trust any single person. Everybody’s fallible; a healthy informational network tends to discover people’s mistakes and point them out. This puts an upper ceiling on how much you can trust even your most beloved leader

nside an echo chamber, that upper ceiling disappears.

Being caught in an echo chamber is not always the result of laziness or bad faith. Imagine, for instance, that somebody has been raised and educated entirely inside an echo chamber

when the child finally comes into contact with the larger world – say, as a teenager – the echo chamber’s worldview is firmly in place. That teenager will distrust all sources outside her echo chamber, and she will have gotten there by following normal procedures for trust and learning.

It certainly seems like our teenager is behaving reasonably. She could be going about her intellectual life in perfectly good faith. She might be intellectually voracious, seeking out new sources, investigating them, and evaluating them using what she already knows.

The worry is that she’s intellectually trapped. Her earnest attempts at intellectual investigation are led astray by her upbringing and the social structure in which she is embedded.

Echo chambers might function like addiction, under certain accounts. It might be irrational to become addicted, but all it takes is a momentary lapse – once you’re addicted, your internal landscape is sufficiently rearranged such that it’s rational to continue with your addiction

Similarly, all it takes to enter an echo chamber is a momentary lapse of intellectual vigilance. Once you’re in, the echo chamber’s belief systems function as a trap, making future acts of intellectual vigilance only reinforce the echo chamber’s worldview.

There is at least one possible escape route, however. Notice that the logic of the echo chamber depends on the order in which we encounter the evidence. An echo chamber can bring our teenager to discredit outside beliefs precisely because she encountered the echo chamber’s claims first. Imagine a counterpart to our teenager who was raised outside of the echo chamber and exposed to a wide range of beliefs. Our free-range counterpart would, when she encounters that same echo chamber, likely see its many flaws

Those caught in an echo chamber are giving far too much weight to the evidence they encounter first, just because it’s first. Rationally, they should reconsider their beliefs without that arbitrary preference. But how does one enforce such informational a-historicity?

The escape route is a modified version of René Descartes’s infamous method.

Meditations on First Philosophy (1641). He had come to realise that many of the beliefs he had acquired in his early life were false. But early beliefs lead to all sorts of other beliefs, and any early falsehoods he’d accepted had surely infected the rest of his belief system.

The only solution, thought Descartes, was to throw all his beliefs away and start over again from scratch.

He could start over, trusting nothing and no one except those things that he could be entirely certain of, and stamping out those sneaky falsehoods once and for all. Let’s call this the Cartesian epistemic reboot.

Notice how close Descartes’s problem is to our hapless teenager’s, and how useful the solution might be. Our teenager, like Descartes, has problematic beliefs acquired in early childhood. These beliefs have infected outwards, infesting that teenager’s whole belief system. Our teenager, too, needs to throw everything away, and start over again.

Let’s call the modernised version of Descartes’s methodology the social-epistemic reboot.

when she starts from scratch, we won’t demand that she trust only what she’s absolutely certain of, nor will we demand that she go it alone

For the social reboot, she can proceed, after throwing everything away, in an utterly mundane way – trusting her senses, trusting others. But she must begin afresh socially – she must reconsider all possible sources of information with a presumptively equanimous eye. She must take the posture of a cognitive newborn, open and equally trusting to all outside sources

we’re not asking people to change their basic methods for learning about the world. They are permitted to trust, and trust freely. But after the social reboot, that trust will not be narrowly confined and deeply conditioned by the particular people they happened to be raised by.

Such a profound deep-cleanse of one’s whole belief system seems to be what’s actually required to escape. Look at the many stories of people leaving cults and echo chambers

Take, for example, the story of Derek Black in Florida – raised by a neo-Nazi father, and groomed from childhood to be a neo-Nazi leader. Black left the movement by, basically, performing a social reboot. He completely abandoned everything he’d believed in, and spent years building a new belief system from scratch. He immersed himself broadly and open-mindedly in everything he’d missed – pop culture, Arabic literature, the mainstream media, rap – all with an overall attitude of generosity and trust.

It was the project of years and a major act of self-reconstruction, but those extraordinary lengths might just be what’s actually required to undo the effects of an echo-chambered upbringing.

we need to attack the root, the systems of discredit themselves, and restore trust in some outside voices.

Stories of actual escapes from echo chambers often turn on particular encounters – moments when the echo-chambered individual starts to trust somebody on the outside.

Black’s is case in point. By high school, he was already something of a star on neo-Nazi media, with his own radio talk-show. He went on to college, openly neo-Nazi, and was shunned by almost every other student in his community college. But then Matthew Stevenson, a Jewish fellow undergraduate, started inviting Black to Stevenson’s Shabbat dinners. In Black’s telling, Stevenson was unfailingly kind, open and generous, and slowly earned Black’s trust. This was the seed, says Black, that led to a massive intellectual upheaval – a slow-dawning realisation of the depths to which he had been misled

Similarly, accounts of people leaving echo-chambered homophobia rarely involve them encountering some institutionally reported fact. Rather, they tend to revolve around personal encounters – a child, a family member, a close friend coming out.

hese encounters matter because a personal connection comes with a substantial store of trust.

We don’t simply trust people as educated experts in a field – we rely on their goodwill. And this is why trust, rather than mere reliability, is the key concept

goodwill is a general feature of a person’s character. If I demonstrate goodwill in action, then you have some reason to think that I also have goodwill in matters of thought and knowledge.

f one can demonstrate goodwill to an echo-chambered member – as Stevenson did with Black – then perhaps one can start to pierce that echo chamber.

the path I’m describing is a winding, narrow and fragile one. There is no guarantee that such trust can be established, and no clear path to its being established systematically.

what we’ve found here isn’t an escape route at all. It depends on the intervention of another. This path is not even one an echo-chamber member can trigger on her own; it is only a whisper-thin hope for rescue from the outside.

Musk’s concern with bots isn’t only a bullshit tactic he’s using to snake out of a bad business deal and/or get a better price for Twitter; it’s also a great example of his shallow thinking. The man has at least some ability to oversee complex engineering systems that land rockets, but his narcissism affords him a two-dimensional understanding of the way information travels across social media.

He is drawn to the conspiratorial nature of bots and information manipulation, because it is a more exciting and easier-to-understand solution to more complex or uncomfortable problems. Instead of facing the reality that many people dislike him as a result of his personality, behavior, politics, or shitty management style, he blames bots. Rather than try to understand the gnarly mechanics and hard-to-solve problems of democratized speech, he sorts them into overly simplified boxes like censorship and spam and then casts himself as the crusading hero who can fix it all. But he can’t and won’t, because he doesn’t care enough to find the answers.

Musk isn’t playing chess or even checkers. He’s just the richest man in the world, bored, mad, and posting like your great-uncle.

a failure to seriously consider the implications of an invention as it becomes entrenched in society; a deep trust in networks, in the idea that allowing people to more freely associate would redound mainly to the good of society; and too much affection for the culture of Silicon Valley and the idea that the people who created a certain thing must have some clue about what to do with it.

I got carried away by the excitement of new tech.

Social networks, I observed, got better as more people used them; it seemed reasonable that at some point one social network would gain widespread acceptance and become a comprehensive directory for connecting everyone.

As an immigrant, I’d also bought into the world-shrinking implications of such a network.

I didn’t consider the far-reaching implications of Facebook’s ubiquity.

What I’d failed to consider was how all these various new things would interact with one another, especially as more people got online.

in calling for everyone to get on Facebook, I should have made a better stab at guessing what could go wrong if we all did. What would be the implications for privacy if we were all using Facebook on our phones — how much could this one service glean about you by being in your pocket all the time?

What would the implications for speech and media be if this single company became a central clearinghouse in the global discourse?

I trusted techies.

This was the vibe pervading media and politics in the late 2000s: Wall Street had ruined the world. Silicon Valley could put it right.

It does not seem in any way good for society — for the economy, for politics, for a basic sense of equality — that a handful of hundred-billion-dollar or even trillion-dollar companies should control such large swathes of the internet.

Obama’s regulators allowed Facebook to buy up its biggest competitors — first Instagram, then WhatsApp — and failed to crack down on its recklessness with users’ private data

“I got caught up in what had become at that point, an extended, combative exchange about policies and procedures,” Gay said in the interview. “What I should have had the presence of mind to do in that moment was return to my guiding truth, which is that calls for violence against our Jewish community — threats to our Jewish students — have no place at Harvard, and will never go unchallenged.”

It was about a widespread culture of killing. It was about prominent white citizens who paid for killings, doctors who administered poisons, morticians who ignored evidence of bullet wounds, lawmen and prosecutors who were on the take and many others who remained complicit in their silence — all because they were profiting from what they referred to openly as the “Indian business.” The real death toll was undoubtedly higher than 24. One bureau agent admitted: “There are so many of these murder cases. There are hundreds and hundreds.”

The movement to suppress elements of American history extends well beyond Oklahoma. According to an analysis by The Washington Post, more than two dozen states have adopted laws that make it easier to remove books from school libraries and to prevent certain teaching on race, gender and sexuality

In 2023, PEN America, which defends freedom of speech, reported that book bans in U.S. public school classrooms and libraries had surged 33 percent over the previous school year, with more than 3,000 recorded removals; among them are classics by the Nobel laureate Toni Morrison (banned in 30 school districts) and Margaret Atwood (banned in 34). School curriculums are being revised to mask discomfiting truths — so much so that in Florida students will now be taught that some African Americans benefited from slavery because it gave them “skills.”

For the well-being of its users, Bejar argued, Meta needed to change course, focusing less on a flawed system of rules-based policing and more on addressing such bad experiences

The company would need to collect data on what upset users and then work to combat the source of it, nudging those who made others uncomfortable to improve their behavior and isolating communities of users who deliberately sought to harm others.

“I am appealing to you because I believe that working this way will require a culture shift,” Bejar wrote to Zuckerberg—the company would have to acknowledge that its existing approach to governing Facebook and Instagram wasn’t working.

During and after Bejar’s time as a consultant, Meta spokesman Andy Stone said, the company has rolled out several product features meant to address some of the Well-Being Team’s findings. Those features include warnings to users before they post comments that Meta’s automated systems flag as potentially offensive, and reminders to be kind when sending direct messages to users like content creators who receive a large volume of messages. 

Meta’s classifiers were reliable enough to remove only a low single-digit percentage of hate speech with any degree of precision.

Bejar was floored—all the more so when he learned that virtually all of his daughter’s friends had been subjected to similar harassment. “DTF?” a user they’d never met would ask, using shorthand for a vulgar proposition. Instagram acted so rarely on reports of such behavior that the girls no longer bothered reporting them. 

Meta’s own statistics suggested that big problems didn’t exist. 

Meta had come to approach governing user behavior as an overwhelmingly automated process. Engineers would compile data sets of unacceptable content—things like terrorism, pornography, bullying or “excessive gore”—and then train machine-learning models to screen future content for similar material.

While users could still flag things that upset them, Meta shifted resources away from reviewing them. To discourage users from filing reports, internal documents from 2019 show, Meta added steps to the reporting process. Meta said the changes were meant to discourage frivolous reports and educate users about platform rules. 

The outperformance of Meta’s automated enforcement relied on what Bejar considered two sleights of hand. The systems didn’t catch anywhere near the majority of banned content—only the majority of what the company ultimately removed

“Please don’t talk about my underage tits,” Bejar’s daughter shot back before reporting his comment to Instagram. A few days later, the platform got back to her: The insult didn’t violate its community guidelines.

Also buttressing Meta’s statistics were rules written narrowly enough to ban only unambiguously vile material. Meta’s rules didn’t clearly prohibit adults from flooding the comments section on a teenager’s posts with kiss emojis or posting pictures of kids in their underwear, inviting their followers to “see more” in a private Facebook Messenger group. 

“Mark personally values freedom of expression first and foremost and would say this is a feature and not a bug,” Rosen responded

Narrow rules and unreliable automated enforcement systems left a lot of room for bad behavior—but they made the company’s child-safety statistics look pretty good according to Meta’s metric of choice: prevalence.

Defined as the percentage of content viewed worldwide that explicitly violates a Meta rule, prevalence was the company’s preferred measuring stick for the problems users experienced.

According to prevalence, child exploitation was so rare on the platform that it couldn’t be reliably estimated, less than 0.05%, the threshold for functional measurement. Content deemed to encourage self-harm, such as eating disorders, was just as minimal, and rule violations for bullying and harassment occurred in just eight of 10,000 views. 

“There’s a grading-your-own-homework problem,”

Meta defines what constitutes harmful content, so it shapes the discussion of how successful it is at dealing with it.”

It could reconsider its AI-generated “beauty filters,” which internal research suggested made both the people who used them and those who viewed the images more self-critical

the team built a new questionnaire called BEEF, short for “Bad Emotional Experience Feedback.

A recurring survey of issues 238,000 users had experienced over the past seven days, the effort identified problems with prevalence from the start: Users were 100 times more likely to tell Instagram they’d witnessed bullying in the last week than Meta’s bullying-prevalence statistics indicated they should.

“People feel like they’re having a bad experience or they don’t,” one presentation on BEEF noted. “Their perception isn’t constrained by policy.

they seemed particularly common among teens on Instagram.

Among users under the age of 16, 26% recalled having a bad experience in the last week due to witnessing hostility against someone based on their race, religion or identity

More than a fifth felt worse about themselves after viewing others’ posts, and 13% had experienced unwanted sexual advances in the past seven days. 

The vast gap between the low prevalence of content deemed problematic in the company’s own statistics and what users told the company they experienced suggested that Meta’s definitions were off, Bejar argued

To minimize content that teenagers told researchers made them feel bad about themselves, Instagram could cap how much beauty- and fashion-influencer content users saw.

Proving to Meta’s leadership that the company’s prevalence metrics were missing the point was going to require data the company didn’t have. So Bejar and a group of staffers from the Well-Being Team started collecting it

And it could build ways for users to report unwanted contacts, the first step to figuring out how to discourage them.

One experiment run in response to BEEF data showed that when users were notified that their comment or post had upset people who saw it, they often deleted it of their own accord. “Even if you don’t mandate behaviors,” said Krieger, “you can at least send signals about what behaviors aren’t welcome.”

But among the ranks of Meta’s senior middle management, Bejar and Krieger said, BEEF hit a wall. Managers who had made their careers on incrementally improving prevalence statistics weren’t receptive to the suggestion that the approach wasn’t working. 

After three decades in Silicon Valley, he understood that members of the company’s C-Suite might not appreciate a damning appraisal of the safety risks young users faced from its product—especially one citing the company’s own data. 

“This was the email that my entire career in tech trained me not to send,” he says. “But a part of me was still hoping they just didn’t know.”

“Policy enforcement is analogous to the police,” he wrote in the email Oct. 5, 2021—arguing that it’s essential to respond to crime, but that it’s not what makes a community safe. Meta had an opportunity to do right by its users and take on a problem that Bejar believed was almost certainly industrywide.

fter Haugen’s airing of internal research, Meta had cracked down on the distribution of anything that would, if leaked, cause further reputational damage. With executives privately asserting that the company’s research division harbored a fifth column of detractors, Meta was formalizing a raft of new rules for employees’ internal communication.

Among the mandates for achieving “Narrative Excellence,” as the company called it, was to keep research data tight and never assert a moral or legal duty to fix a problem.

“I had to write about it as a hypothetical,” Bejar said. Rather than acknowledging that Instagram’s survey data showed that teens regularly faced unwanted sexual advances, the memo merely suggested how Instagram might help teens if they faced such a problem.

The hope that the team’s work would continue didn’t last. The company stopped conducting the specific survey behind BEEF, then laid off most everyone who’d worked on it as part of what Zuckerberg called Meta’s “year of efficiency.

If Meta was to change, Bejar told the Journal, the effort would have to come from the outside. He began consulting with a coalition of state attorneys general who filed suit against the company late last month, alleging that the company had built its products to maximize engagement at the expense of young users’ physical and mental health. Bejar also got in touch with members of Congress about where he believes the company’s user-safety efforts fell short. 

Pandu Nayak, a vice president for search who leads the company’s search-quality teams, told me that snippets are designed to be helpful to the user, to surface relevant and high-caliber results. He argued that they are “usually an invitation to learn more” about a subject

“It’s a strange world where these massive companies think they’re just going to slap this generative slop at the top of search results and expect that they’re going to maintain quality of the experience,” Nicholas Diakopoulos, a professor of communication studies and computer science at Northwestern University, told me. “I’ve caught myself starting to read the generative results, and then I stop myself halfway through. I’m like, Wait, Nick. You can’t trust this.”

Nayak said the team focuses on the bigger underlying problem, and whether its algorithm can be trained to address it.

If Nayak is right, and people do still follow links even when presented with a snippet, anyone who wants to gain clicks or money through search has an incentive to capitalize on that—perhaps even by flooding the zone with AI-written content.

Nayak told me that Google plans to fight AI-generated spam as aggressively as it fights regular spam, and claimed that the company keeps about 99 percent of spam out of search results.

The result is a world that feels more confused, not less, as a result of new technology.

The Kenya result still pops up on Google, despite viral posts about it. This is a strategic choice, not an error. If a snippet violates Google policy (for example, if it includes hate speech) the company manually intervenes and suppresses it, Nayak said. However, if the snippet is untrue but doesn’t violate any policy or cause harm, the company will not intervene.

experts I spoke with had several ideas for how tech companies might mitigate the potential harms of relying on AI in search

For starters, tech companies could become more transparent about generative AI. Diakopoulos suggested that they could publish information about the quality of facts provided when people ask questions about important topics

They can use a coding technique known as “retrieval-augmented generation,” or RAG, which instructs the bot to cross-check its answer with what is published elsewhere, essentially helping it self-fact-check. (A spokesperson for Google said the company uses similar techniques to improve its output.) They could open up their tools to researchers to stress-test it. Or they could add more human oversight to their outputs, maybe investing in fact-checking efforts.

Fact-checking, however, is a fraught proposition. In January, Google’s parent company, Alphabet, laid off roughly 6 percent of its workers, and last month, the company cut at least 40 jobs in its Google News division. This is the team that, in the past, has worked with professional fact-checking organizations to add fact-checks into search results

Alex Heath, at The Verge, reported that top leaders were among those laid off, and Google declined to give me more information. It certainly suggests that Google is not investing more in its fact-checking partnerships as it builds its generative-AI tool.

Nayak acknowledged how daunting a task human-based fact-checking is for a platform of Google’s extraordinary scale. Fifteen percent of daily searches are ones the search engine hasn’t seen before, Nayak told me. “With this kind of scale and this kind of novelty, there’s no sense in which we can manually curate results.”

Creating an infinite, largely automated, and still accurate encyclopedia seems impossible. And yet that seems to be the strategic direction Google is taking.

A representative for Google told me that this was an example of a “false premise” search, a type that is known to trip up the algorithm. If she were trying to date me, she argued, she wouldn’t just stop at the AI-generated response given by the search engine, but would click the link to fact-check it.

This pervasive claim that “I have my rights” but “I don’t have responsibilities” is unraveling our country today.

But societal immunity “is a function of trust,”

“When trust in institutions, leaders and each other is high, people — in a crisis — are more willing to sublimate their cherished rights and demonstrate their sense of shared responsibilities toward others, even others they disagree with on important issues and even if it means making sacrifices.”

When our trust in each other erodes, though, as is happening in America today, fewer people think they have responsibilities to the other — only rights that protect them from being told by the other what to do.

completely ignored the four most important statistical facts about Covid-19 today that highlight our responsibilities — to our fellow citizens and, even more so, to the nurses and doctors risking their lives to take care of us in a pandemic.

First, unvaccinated adults 18 years and older are 16 times more likely to be hospitalized for Covid than fully vaccinated adults

Second: Adults 65 and older who are not vaccinated are around 50 times more likely to be hospitalized for Covid than those who have received a full vaccine course and a booster.

Third: Unvaccinated people are 20 times more likely to die of Covid than people who are vaccinated and boosted.

the emotional toll and other work conditions brought on by the pandemic contributed to some two-thirds of nurses giving thought to leaving the profession.

many hospitals today are experiencing an unprecedented 20 percent annual turnover rate of nurses — more than double the historical baseline. The more nurses leave, the more those left behind have had to work overtime.

Especially when so many dying unvaccinated patients tell their nurses, “I wish I had gotten vaccinated,”

none of these statistics were mentioned during that podcast

“You can listen to the entire 186-minute lovefest between Rogan and Malone and have no idea that our hospitals are overloaded with Covid cases,” wrote Levy, “and that on the day their conversation transpired, 7,559 people worldwide died of Covid, 1,410 of which were in the United States. The vast majority of them were unvaccinated.”

“When Malone uncorks questionable allegations about disastrous vaccine effects and the global cabal of politicians and drugmakers pulling strings, Rogan responds with uh-huhs and wows.”

That was Rogan’s right. That was Spotify C.E.O. Daniel Ek’s right.

But who was looking out for the doctors and nurses on the pandemic front lines whose only ask is that the politicians and media influencers who are privileged enough to have public platforms — especially one like Rogan with an average of 11 million listeners per episode — use them to reinforce our responsibilities to one another, not just our rights.

He could start by offering his listeners a 186-minute episode with intensive care nurses and doctors about what this pandemic of the unvaccinated has done to them.

That would be a teaching moment, not only about Covid, but also about putting our responsibilities to one another — and especially to those who care for us — at least on a par with our right to be as dumb and selfish as we want to be.

hose promises have also fueled anxiety over how people will be able to compete for jobs outsourced to eerily refined machines or trust the accuracy of what they see online.

Officials with the San Francisco lab said GPT-4’s “multimodal” training across text and images would allow it to escape the chat box and more fully emulate a world of color and imagery, surpassing ChatGPT in its “advanced reasoning capabilities.”

A person could upload an image and GPT-4 could caption it for them, describing the objects and scene.

AI language models often confidently offer wrong answers because they are designed to spit out cogent phrases, not actual facts. And because they have been trained on internet text and imagery, they have also learned to emulate human biases of race, gender, religion and class.

GPT-4 still makes many of the errors of previous versions, including “hallucinating” nonsense, perpetuating social biases and offering bad advice. It also lacks knowledge of events that happened after about September 2021, when its training data was finalized, and “does not learn from its experience,” limiting people’s ability to teach it new things.

Microsoft has invested billions of dollars in OpenAI in the hope its technology will become a secret weapon for its workplace software, search engine and other online ambitions. It has marketed the technology as a super-efficient companion that can handle mindless work and free people for creative pursuits, helping one software developer to do the work of an entire team or allowing a mom-and-pop shop to design a professional advertising campaign without outside help.

it could lead to business models and creative ventures no one can predict.

sparked criticism that the companies are rushing to exploit an untested, unregulated and unpredictable technology that could deceive people, undermine artists’ work and lead to real-world harm.

the company held back the feature to better understand potential risks. As one example, she said, the model might be able to look at an image of a big group of people and offer up known information about them, including their identities — a possible facial recognition use case that could be used for mass surveillance.

OpenAI researchers wrote, “As GPT-4 and AI systems like it are adopted more widely,” they “will have even greater potential to reinforce entire ideologies, worldviews, truths and untruths, and to cement them or lock them in.”

“We can agree as a society broadly on some harms that a model should not contribute to,” such as building a nuclear bomb or generating child sexual abuse material, she said. “But many harms are nuanced and primarily affect marginalized groups,” she added, and those harmful biases, especially across other languages, “cannot be a secondary consideration in performance.”

OpenAI said its new model would be able to handle more than 25,000 words of text, a leap forward that could facilitate longer conversations and allow for the searching and analysis of long documents.

OpenAI developers said GPT-4 was more likely to provide factual responses and less likely to refuse harmless requests

Duolingo, the language learning app, has already used GPT-4 to introduce new features, such as an AI conversation partner and a tool that tells users why an answer was incorrect.

The company did not share evaluations around bias that have become increasingly common after pressure from AI ethicists.

GPT-4 will have competition in the growing field of multisensory AI. DeepMind, an AI firm owned by Google’s parent company Alphabet, last year released a “generalist” model named Gato that can describe images and play video games. And Google this month released a multimodal system, PaLM-E, that folded AI vision and language expertise into a one-armed robot on wheels: If someone told it to go fetch some chips, for instance, it could comprehend the request, wheel over to a drawer and choose the right bag.

The systems, though — as critics and the AI researchers are quick to point out — are merely repeating patterns and associations found in their training data without a clear understanding of what it’s saying or when it’s wrong.

GPT-4, the fourth “generative pre-trained transformer” since OpenAI’s first release in 2018, relies on a breakthrough neural-network technique in 2017 known as the transformer that rapidly advanced how AI systems can analyze patterns in human speech and imagery.

The systems are “pre-trained” by analyzing trillions of words and images taken from across the internet: news articles, restaurant reviews and message-board arguments; memes, family photos and works of art.

Giant supercomputer clusters of graphics processing chips are mapped out their statistical patterns — learning which words tended to follow each other in phrases, for instance — so that the AI can mimic those patterns, automatically crafting long passages of text or detailed images, one word or pixel at a time.

In 2019, the company refused to publicly release GPT-2, saying it was so good they were concerned about the “malicious applications” of its use, from automated spam avalanches to mass impersonation and disinformation campaigns.

Altman has also marketed OpenAI’s vision with the aura of science fiction come to life. In a blog post last month, he said the company was planning for ways to ensure that “all of humanity” benefits from “artificial general intelligence,” or AGI — an industry term for the still-fantastical idea of an AI superintelligence that is generally as smart as, or smarter than, the humans themselves.

In an Atlantic essay, “Why The Past Ten Years of American Life have Been Uniquely Stupid,” and in a follow-up piece, Haidt argued that the Age of Gutenberg–of books and the depth understanding that comes with them–ended somewhere around 2014 with the rise of “Share,” “Like” and “Retweet” buttons that opened the way for trolls, hucksters and Trumpists

The new age of “hyper-virality,” he writes, has given us both January 6 and cancel culture–ugly polarization in both directions. On the subject of stupidification, we should add the fact that high school students now get virtually their entire stock of knowledge about the world from digital platforms.

Haidt proposed several reforms, including modifying Facebook’s “Share” function and requiring “user verification” to get rid of trolls. But he doesn’t really believe in his own medicine

Haidt said that the era of “shared understanding” is over–forever. When I asked if he could envision changes that would help protect democracy, Haidt quoted Goldfinger: “Do you expect me to talk?” “No, Mr. Bond, I expect you to die!”

Social media is a public health hazard–the cognitive equivalent of tobacco and sugary drinks. Adopting a public health model, we could, for examople, ban the use of algorithms to reduce virality, or even require social media platforms to adopt a subscription rather than advertising revenue model and thus remove their incentive to amass ev er more eyeballs.

We could, but we won’t, because unlike other public health hazards, digital platforms are forms of speech. Fox New is probably responsible for more polarization than all social media put together, but the federal government could not compel it–and all other media firms–to change its revenue model.

If Mark Zuckerberg or Elon Musk won’t do so out of concern for the public good–a pretty safe bet–they could be compelled to do so only by public or competitive pressure. 

Taiwan has provide resilient because its society is resilient; people reject China’s lies. We, here, don’t lack for fact-checkers, but rather for people willing to believe them. The problem is not the technology, but ourselves.

you have to wonder if people really are repelled by our poisonous discourse, or by the hailstorm of disinformation, or if they just want to live comfortably inside their own bubble, and not somebody else’

If Jonathan Haidt is right, it’s not because we’ve created a self-replicating machine that is destined to annihilate reason; it’s because we are the self-replicating machine.

It is reasonable to imagine that Rife thinks his audience processes his Black English usage as a warm method of interpersonal bonding in the same way he seems to. In fact, a tweet of his suggests that he hadn’t even been conscious of what he was doing until apprised, and doesn’t even think of himself as shifting into something “Black” at all.

Rife is not posing or ridiculing; he’s connecting. Linguists call it accommodation. A non-Black speaker these days may do it with a Black audience.

reflecting in language exactly what we were supposed to be going for: interracial harmony.

Except that these days, that ideal may seem a tad 1.0. Under the new identitarian mind-set, where we cherish coming together less than we cherish a diversity of identities, many see someone like Rife as culturally appropriating Black speech, something that isn’t his. “Mimesis is a kind of negation,”

There is simply no way that whiteness and Blackness will mingle as they have in music, cuisine, gesture, greeting styles, dating, matrimony and multiracial identity, and yet for some reason be halted at language.

The horse has been out of the barn ever since white kids embraced Jay-Z and Tupac.

style-shifting is humans’ linguistic default, not a pose or party trick.