Bob Metcalfe, The Man Who Discovered Network Effects, Isn’t Sorry
Get real time updates directly on you device, subscribe now.
ChatGPT warned me against asking legendary engineer Bob Metcalfe about his 1996 prediction that the internet would collapse. This came after I sought the chatbot’s guidance on what questions to ask the man who this week received the ACM Turing Award, the $1 million prize dubbed the Nobel of computing. The AI oracle suggested I stick to quizzing him on his famous accomplishments—inventing Ethernet, starting the 3Com Corporation, codifying the value of networks, and teaching students in Texas about innovation, which he did until he retired last year “to pursue a sixth career.”
But ChatGPT thought it was a terrible idea to bring up Metcalfe’s bold prognostication, just as the network he’d helped pioneer was taking off, that the volume of bits zipping around the internet would cause the mother of all crashes. OpenAI’s black box told me that since Metcalfe’s guess had flopped in a very public manner, I’d be risking the honoree’s pique if I raised it, and from then on he’d be too annoyed to share his best thoughts. The interview would be a disaster.
Oh-kay, I thought. And then I clicked on the Zoom link.
The prizewinner who greeted me looked terrific at 76, hardly changed from the guy I last saw maybe 30 years ago when he was running tech conferences and hosting great parties at his mansion in Boston’s Back Bay. (He spoke to me from his home in Austin, where he had moved for his teaching gig.)
For someone known for his bluster, he seemed genuinely humbled to join the Turing club, though you might say it took them long enough. It was almost 50 years ago to the day that Metcalfe wrote a memo to his bosses at Xerox Palo Alto Research Center proposing a way to connect the lab’s innovative personal computers to its groundbreaking laser printer, and to one another. Inspired by an obscure Hawaiian system called AlohaNet, he figured out a way to dynamically handle high-speed data in a network without having the bits clash or forcing reconfiguration each time a new user showed up. He dubbed it Ethernet. (He developed it with a co-inventor, David Boggs.)
Metcalfe’s idea not only solved the problem at PARC, but wound up scaling into a vital technology for everyone. Over 5 billion people use the internet. Did he have that in mind when he concocted those first networks? “No, although it’d be convenient for me to say so,” he says. “PARC was a very much ‘build your own tools’ kind of place. But in retrospect, what we were doing was helping the internet transition from the networking of dumb terminals to the networking of personal computers.”
In 1979, Metcalfe founded 3Com to help commercialize Ethernet, after he’d persuaded Xerox to make the networking technology an open standard. Throughout the 1980s he relentlessly promoted the standard; by then he’d made a brilliant observation that explained the growth of not just the internet, but also the many services built on top of it: that the value of a network is proportional to the square of the number of users. In other words, each time a new user joins a network it grows more powerful.
In 1985, the economist George Gilder named the idea Metcalfe’s law. It’s probably the most celebrated equation of its kind since Gordon Moore’s observation about computer chips. Metcalfe says his motivation was not science but commerce. “It was a sales tool,” he says. “People were building small networks and not finding them useful. So I ginned up a slide on an Alto that showed that the cost of a network goes up linearly with the number of nodes, but the number of possible connections goes up as the square. Our salesforce took this 35-millimeter slide and told people the reason they weren’t useful is that they weren’t big enough. The remedy, of course, was buying more of our networks.”
When people talk of network effects, they’re unwittingly channeling that sales pitch. You can argue that the idea was behind the explosion of not just social networks, but the entire philosophy of “get big fast” that characterized the first two decades of tech in this century. So you would think that the day Bob Metcalfe actually visited Facebook, he would have been treated like a hero. His daughter, an early Facebook employee, had arranged a meeting with CEO Mark Zuckerberg and his right-hand woman Sheryl Sandberg, so that the executives could connect with the man whose achievements made their company possible. It didn’t work out so well. “Mark stood me up and Sheryl didn’t know what Metcalfe’s law was,” he says. “She invited in a Stanford PhD who she thought might know something about Metcalfe’s law. And he didn’t either. So the conversation was short.”
Metcalfe acknowledges that the power of networks isn’t always to the good. “We don’t quite know how to handle all the connectivity,” he says. “That’s why there are these pathologies emanating. But I am confident that we’re going to figure out how to handle truth, so that, you know, fake news doesn’t kill us.” After all, he notes, at one point people thought porn was going to ruin the internet. “But now we don’t argue about porn anymore, that’s been handled. Even advertising was viewed as a pathology for a while—until it funded the entire internet.” (We didn’t pause to debate the fact that porn is still an issue and that some people think online ads ruined civilization.)
Speaking of the internet, I defied ChatGPT’s advice and reminded Metcalfe of the 1996 collapse he predicted that never happened. If the chatbot had truly understood Metcalfe, it would have known that he’d greet the subject with a laugh. He recounted that he’d been so confident of a crash that he had promised to eat the column if he was wrong. And so at a 1997 Web conference, after checking that the ink wouldn’t be toxic, he put the newsprint in a blender with some unidentified liquid and drank it all, before a delighted audience.
Now that Metcalfe has a Turing, he can gulp down some champagne instead of his own words, and get back to his so-called sixth career: using computational engineering to improve geothermal energy production. He wisely offers no predictions on how that will work out.
In September 1996, I wrote about Metcalfe’s prediction of internet collapse for Newsweek. Twenty-seven years later, the only thing that’s collapsed is the uppercase in the internet’s first letter. Capitalization is no longer necessary because the word “internet” is now officially a common noun as opposed to a proper one, as common as air or water. Even though Newsweek has changed hands a few times since, you can still find the column on the good old internet.
In [Metcalfe’s] thinking, the Net is a fish already hooked; those routine brownouts are but the first few twinges at its mouth. Soon the fish will find itself reeled in, and we will witness the pathetic spectacle of the once mighty Internet, the darling of our economy and the object of our millennial dreams, flopping aimlessly like snagged red snapper on a boat deck. Web sites will become cobweb sites. E-mail will be dead-lettered. Stock prices will fall to Earth.
“Maybe the Internet has already collapsed,” Metcalfe says to me later that day, in a lecture that lasts the dinner hour. “Everyone complains about brownouts every day. But it’s going to get worse. Worse! Worse! Worse!” The inventor of the Ethernet networking system and founder of the 3Com corporation has risked his considerable reputation by publicly predicting a “gigalapse”—in which a billion hours of access time are lost—by the end of the year. This would far eclipse the recent dead spells at America Online and Netcom, one of the biggest Internet service providers. More important, the collapse could also cast doubts on the reliability of cyberspace itself. Since the consensus is that just about everything from movies to mail is eventually going to move to the Internet, this is serious stuff.
The problem, as Metcalfe sees it, is that the Net is incapable of handling the millions of immigrants washing up on its virtual shores. There is neither the physical capacity to shuttle all of those bits of information its users generate nor the organizational capacity to address the problem. Some people rhapsodize about the fact that the Internet has no president, no police and no blueprints for organized growth; new standards are arrived at by consensus, and it’s up to companies like Sprint and MCI to beef up the “backbone” that handles the bulk of the information flow. To Metcalfe this is more travesty than rhapsody. He rails against the destructive “ideology” that celebrates the decentralized, semi-anarchic structure of the Internet. Metcalfe thinks it’s time for engineers in sandals to step aside and let professionals in suits run the show.
Gabor asks, “What share of news and newsletters will be written by AI Large Language Models in a year from now? How will that change your industry?”
Thanks, Gabor (I think). The ability of chatbots made by OpenAI, Google, and others to churn out readable copy has already led people to jam editorial transoms with instant articles. They are competent but not inspired; still, for certain kinds of stories—particularly those that depend on summarizing existent content—these instant stories can be “good enough.” One baseline that must be honored: fact-checking. Because the bots tend to make things up. These are called “hallucinations.” That might have worked for Hunter S. Thompson but the practice is discouraged in much of journalism.
So, yes, a lot of content at the bottom of the news food chain—the one that depends on raw information rather than a sophisticated take on what’s happening—will probably be proliferating via algorithms a year from now. But the things you pay for—like deeply reported news by a writer whose perspective is informed by knowing how the world actually works—seem safe for now. I would expect, however, that LLMs’ ability to gather and organize vast amounts of information almost instantly might become a useful tool for those reporters. So one big industry change will be a routine collaboration with AI. (Here’s WIRED’s new policy on that.)
But don’t mistake today’s LLMs for tomorrow’s. Looking past your one-year time horizon, there will be dramatic improvement that might force even bigger changes. How much this threatens the top of the news quality pyramid is an open question. A good news article requires shoe-leather reporting. Can a bot do that? Will news sources respond to an interview request from GPT-7? And can chatbot designers figure out how to get rid of the hallucinations, or keep just enough to incubate the next great gonzo journalist?
You can submit questions to mail@wired.com. Write ASK LEVY in the subject line.
“The climate time bomb is ticking,” says a new report, and only immediate, drastic measures can defuse it. The US, China, and other great powers immediately put those necessary measures into effect. Just kidding, they did nothing and we’re doomed.
Our writer went to a Taiwanese semiconductor factory and saw the face of God. No wonder Biden’s spending all that dough to make chips here.
One more thing to worry about: feral hogs.
How tech founders with big deposits in Silicon Valley Bank got through the Worst Weekend Ever.
The Russian space program is in trouble. Paging Elonavich Muskofsky.
