In early November, as mass layoffs began at Twitter and blue-check parodies were about to nuke the company’s advertising business, the new owner got a little dorm-roomy. At 1:13 a.m. on a Thursday, Elon Musk tweeted:
“Because it consists of billions of bidirectional interactions per day, Twitter can be thought of as a collective, cybernetic, super-intelligence.”
This was a pretty wild thing to be thinking about, given all the potentially company-ending chaos that Musk was getting ready to inflict on Twitter. But it also revealed something important about his thought process. Musk has long worried, very publicly, about the dangers that all-powerful cybernetic superintelligences might pose to humanity. So you might think that if Twitter were one of those, that would have been high on his list of reasons not to spend $44 billion to buy the company. Perhaps he didn’t perceive that particular caveat until he was the emptor. Or perhaps the price he paid for Twitter wasn’t the only thing that was high.
Either way, Musk had stumbled on to a rich theoretical vein. Flocks of birds, schools of fish, herds of cattle, swarms of bees, even tumors, brains, and sometimes drones and software agents — what we might collectively call “collectives” — do preternaturally intelligent things when they work in unison.
So Musk was on to something big. Biologists, anthropologists, and information theorists do think that social networks, like Musk’s bird app, show at least some signs of being flocks. On a social network, all the likes and faves, the mutual follows and retweets and shares, turn us individual users into something bigger, smarter, and weirder. And scientists hope that the mechanisms for how that works could someday help tame the crappier aspects of social media — the polarization, the disinformation, the harassment, the Nazis. Understanding Twitter as a collective could make social media less polarizing and more useful.
But the thing is, I don’t think that’s what Musk meant. And since he hit on the idea of Twitter as a collective intelligence, he’s gone on to get every implication of that larger thought wrong. At a core level, he simply doesn’t understand what he bought, or how it works. And whether Musk manages to hold this thing together or spin it into shards, his deeper misunderstanding should make all of us even more worried about the future of social media than we already are. If Twitter is a collective superbrain, that superbrain might be sociopathic.
Resistance is futile
Here’s a chilling sentence: Elon Musk was right.
A group of seemingly random-acting individuals turns into a collective when it follows a set of simple rules, like “turn right when the guy nearest you turns right” or “make an alarmed noise when you hear an alarmed noise.” From these tiny instruction sets, all sorts of complicated cooperative actions (pack hunting, migrations) arise spontaneously, like a weaver creating an intricate twill pattern just by repeating a couple of simple flicks of their loom. Scientists call those emergent behaviors.
For the behaviors to emerge, though, animals have to communicate. Fish and birds use visual signals about what their neighbors are doing — turn left, dive fast, whatever — to create the gorgeous flowing murmurations of starlings or quick en-masse undulations of anchovy schools. Hyenas use audible calls. Ants lay down pheromone trails. And people? We have language. That’s how we trade information.
In that sense, social networks are definitely collectives. Where else do so many humans communicate with so many other humans than on Twitter and Facebook and TikTok? They’re collectives, and things like viral memes or Arab Springs are what emerge. “Collective behavior happens wherever you have rules of interaction among individuals. You get emergent properties,” says Joe Bak-Coleman, a researcher at the Craig Newmark Center for Journalism Ethics and Security at Columbia University who studies this stuff. “But that’s quite different from the question of, at these very large scales, are we processing information and making good decisions?”
Musk’s tweet attracted Bak-Coleman’s attention because he has been working for years on the idea that understanding the collectivist nature of social networks might make them better. Yes, people communicate on social networks. But the twist, Bak-Coleman warns, is that “social networks change how that information propagates.” Our words go further, faster. We don’t get any of the signals of trustworthiness our brains have evolved to look for. If the bird closest to you says “turn right,” he may just be trying to trick you into voting for Donald Trump. And since we’re all more likely to spread signals that are new, surprising, or emotionally fraught, disinformation and rage move faster online than truth or beauty.
But there’s an upside. As Bak-Coleman and a bunch of his colleagues pointed out last year in a paper called “Stewardship of Global Collective Behavior,” all those ideas circulating so widely and rapidly mean that one simple tweak could make a social network a lot more pleasant. All that’s needed is to tap the brakes — to add a little bit more friction to the system, making it fractionally harder for any individual expression to go viral.
In most high-functioning networks, signals start to degrade after just three or four degrees of separation. The lords and masters of Twitter or Facebook could set their systems so that when an idea threatens to infect the internet, circuit breakers activate. Think of it like mask-wearing and ventilation, but for memes. Fresh ideas might take a little longer to spread. But the germs will get filtered out.
The desire for a better filtering system may be why so many Twitter users are decamping to the newish social network Mastodon like ships fleeing a sinking rat. Mastodon’s many servers (or “instances”), each with its own rules of behavior, make wideband communication slightly less easy than on Twitter. That adds up to a kind of “antivirality,” as the tech writer Clive Thompson put it. The network disfavors speed and distance, which makes the overall experience more pleasant.
But Musk isn’t doing any of that. Yes, he figured out what Twitter is. But he failed to grasp how it operates, or why.
The rules aren’t simple
Which brings us to a more familiar sentence: Elon Musk was wrong.
Collectives emerge only when they follow simple rules laid out by physics and biology — when the group itself decides on a course of action. But social networks are built on rails, governed by algorithms, which interferes with the self-organizing. Musk doesn’t seem to understand that Twitter can’t turn into a collective cybergenius if he, the owner, won’t let it have a mind of its own.
“Elon’s tweet is basically espousing the invisible hand of social behavior,” Bak-Coleman says. “We just connect everyone and the invisible hand of collective intelligence will usher in a utopia with free speech and no violence? That would maybe play well on Joe Rogan’s podcast to a stoned listener, but it’s no different than the claim that the economy will just work itself out.”
Just like an economy, a social network has rules and regulations. And the current rules deployed by Twitter and most of the other dominant social networks are designed to subtly enhance conflict. The thermostat is set a little high; the chairs are a little cramped. Why? Because all those algorithmic choices keep us clicking. “They drive up engagement and mine our attention to feed us ads, which we engage with and buy stuff, and that generates revenue for the site,” Bak-Coleman says.
No group can develop collective cognition if all its members are trolling one another. “A brain that’s infighting and unable to come to a consensus would be unable to operate,” says Iain Couzin, an expert in collective behavior who serves as the director of the Max Planck Institute of Animal Behavior at the University of Konstanz. “We do not have natural selection operating on Twitter, so the analogy fails. Facebook, Twitter — all the human social networks do not have that property.”
In this construction, a social network might have become a collective superintelligence, had capitalists left it to its — our? — own devices. But the algorithms and ads that make it profitable exclude the possibility of emergent collective cognition. They murder the superintelligence in its infancy, and in its place we get a dumb machine optimized to extract profit from our attention.
The truth is probably somewhere in between. Maybe Musk was right, but wrong about how he was right. A social network on algorithmic rails could be an emergent uberbrain and also be terrible, a cybernetic superintelligence that will be extremely hardcore in pursuit of profit.
The nature of Twitter’s superintelligence, like Musk’s, remains a matter of faith.
The science of collectivity suggests that social networks come in two flavors: kind and gentle, or profitable. Think of all the different approaches to social media: short video clips, long blog entries, short text, still pictures, moderated, unmoderated, anonymous, and so on. So far, all of the for-profit ones have ended up chaotic-evil. But even the foremost experts in the field will be the first to acknowledge they don’t really know why that is, or how to fix it.
“We have no idea what produces collective intelligence among humans, especially at large scales,” says Duncan Watts, a computational social scientist at the University of Pennsylvania who has worked as a researcher at Yahoo and Microsoft. Finding some science to tame social networks would be great, of course — “a super important question, both for science and society,” Watts says. “But it’s so far removed from what most of social science has actually established that I’m not sure we know anything useful at all.”
Studying collectives is hard, y’all. When the networks are run by public companies and have 100 million users? Forget about it. “We have zero clue,” Bak-Coleman concedes. “Well, not zero clue. But if Elon Musk decided that the current recommendation system is bad, let’s say, and replaced it with a whole new system, and shared the code and data with scientists, and gave us a year and a half with it, we still couldn’t tell you what that would do to democracy.”
Couzin says much the same. “It would be helpful if there was an openness about these algorithms,” he says. “There’s an interesting ethical aspect to the control, the subtle dials they have to control the information structure.”
So what did Musk do the minute he took over Twitter? He fired the team that studies this stuff and shares data with outside researchers. So the nature of Twitter’s superintelligence, like Musk’s, remains a matter of faith.
We really are all in this together, alas
The single most tantalizing piece of evidence supporting the proposition that Twitter is a positive, uplifted, emergent superintelligence comes, perhaps ironically, from what looks to be its final act: the fact that so many people are leaving it.
One of the most basic things a collective does is move, whether it’s bacteria attacking a new organ or elephants walking to water. So the way everyone on Twitter seems to be trudging across the digital veldt toward Mastodon, Couzin says, “is a collective migration, sharing many of the hallmarks we see in animal networks.”
But not just any animal network. Specifically, we’re all acting like honeybees looking for a new nest. When a hive breaks down, a honeybee swarm sends out scouts with the specs for a perfect nest encoded in their brains — things like size of entrance, location, square footage, and so on. Scouts are the realtors of the honeybee world. They find candidate locations and then fly back to the cluster of bees looking for a new home, where they argue their case. By dancing.
Each scout’s choreography encodes the direction and distance to the site it’s touting. The better the site is, the longer and more vigorously the scout dances, recruiting other scouts and non-scout bees to its moves. The scene eventually turns into a whole-hive dance-off — “Step Up: The Swarm” — until only one team is left. All the bees align into a massive apiary Bollywood number that also teaches everyone the new nest’s coordinates. Then they all buzz off.
So: To bee or not to bee? Under Musk, Twitter has entered the dance-off phase. I probably spend too much time on there, but my experience of the past couple of weeks has been that of a bee watching the scouts return. Some groove toward Mastodon; others get down to Cohost. I have even picked up some new dance moves. I’ll be sad if the Twitter superintelligence starts singing a Kubrickian cover of “Daisy” and implodes into a pile of melting isolinear chips. But I look forward to being part of a new collective, wherever we decide to locate our next hive-mind.
The analogy isn’t perfect, of course. And as the scientists are quick to point out, you can’t equate people with insects. “Humans,” Couzin says, “are much more intelligent than bees.”
Uh-huh. Sure we are.
Adam Rogers is a senior correspondent at Insider.