In 1998, children’s author Thacher Hurd published Zoom City, a picture book for toddlers in which friendly animals drive cars around a bustling urban metropolis. Tech investor Jim Scheinman read it to his children, and he still remembered the book 14 years later when he advised a Silicon Valley founder named Eric Yuan. 

“I loved this fun little book as much as my kids, and hoped to use the name someday for the perfect company that embodied the same values of creativity, exploration, happiness, and trust,” Scheinman later wrote. Which is how, with the help of a pandemic multiplier effect, a little-known videoconferencing startup called Saasbee Inc. became Zoom, the all-conquering corporate behemoth that dominates workplace calendars.

The pandemic is officially over, yet Zoom City—the city that Zoom built—endures. But right now it’s not a place of creativity, exploration, happiness, or trust. Instead, remote work has turned some global cities into dried-out urban husks—decaying downtowns full of shuttered sandwich shops and empty office blocks. As businesses leave or downsize, city tax revenues decline, meaning less money to spend on public services. And the people left behind because they can’t work remotely get trapped in the doom loop

This is particularly apparent in San Francisco, which has been dubbed “the most empty downtown in America.” Much of the technology that enabled the transition to remote work emerged from the Bay Area, but it also created a combination of traits—demographics, industry norms, property prices—that has made workers here particularly unlikely to return to the office. 

The New York Times recently reported that office occupancy in SF is at 40 percent of its pre-pandemic level, roughly 7 percentage points below the average major US city. It’s facing a $728 billion budget hole at the same time as grappling with a suite of problems—homelessness, drug abuse, crime—that have been exhaustively well documented (arguably by those with a vested interest in singling out a rich and progressive city).

None of these issues are unique to San Francisco, though. I’m based in London but have spent the past month working from WIRED’s San Francisco office for a change of pace from suburban England—fewer horses, more horse tranquilizer—and as an outsider I have found the descriptions of a postapocalyptic hellscape to be fairly overblown. (This from the UK’s Daily Mail is particularly deranged, as it tries to pin the blame for SF’s problems entirely on working from home).

But walking to work each morning through what is admittedly a pretty quiet downtown has got me thinking about the future of these urban spaces, and what to do with office buildings when we don’t need offices anymore. 

This is a global issue, but it’s one that has particularly affected American cities, says Yonah Freemark, a senior research associate at the Urban Institute, who has been writing eloquently about planning issues for more than a decade. “Downtowns in the United States are uniquely monofunctional in form compared to almost any other part of the world,” he says. Strict zoning laws, combined with the widespread leveling of city centers in the 1960s and 1970s to build multilane highways, have created downtowns that are difficult to use for anything other than white-collar work.

recent op-ed by Harvard economist Edward Glaeser and MIT’s Carlo Ratti argues that we’re entering the era of the “playground city,” where downtown areas will be remodeled to attract leisure visitors as well as workers. I’ve definitely spotted this pattern in London, where people who go to the office only a couple of days a week arrange their work schedules to complement their social calendars, rather than the other way around. It’s similar to what has happened to British high streets over the past 20 years, where retailers decimated by online shopping have been replaced by bars, cafés, and restaurants.

But Freemark is wary of encouraging the development of playground cities, worried that they might become the next fad in urban planning. Before the pandemic, the “smart city” was in vogue, and prior to that it was city malls and festival marketplaces. “The key goal over the short to medium term should be to create public spaces that are appealing, lively, and attractive,” he says. That could be as simple and low-cost as adding a coffee kiosk to a park, or some more benches to a public square. “Create spaces where people want to be,” he says. “It doesn’t have to be an aquarium.”

Freemark lives in Washington, DC, where Mayor Muriel Bowser has pursued a city-wide redesign that aims to ensure, Freemark says, “motorists from the suburbs can get in more easily.” He believes that this is exactly the wrong solution. “My sense is that a city’s goal should be to create neighborhoods that are vibrant and livable,” he says.

One thing that’s true of both London and San Francisco is that while central business districts have struggled, residential neighborhoods have thrived. Former commuter towns are stuffed with workers with newly discovered free time that they really don’t want to use making their own lunch. Places like Mill Valley, across the Golden Gate Bridge from San Francisco, have been attracting new businesses, drawing the same lunch chains that might once have been confined to downtowns.

Perhaps the answer to saving downtowns is actually a simple one: Transform them into neighborhoods in their own right that actually cater to the needs of the people who live there. College students, Freemark suggests, are one group that might benefit from the cheaper living costs and good access to public transportation offered by a wide-scale conversion of downtown buildings into high-density residential dwellings. On WIRED’s Have a Nice Future podcast, San Francisco’s mayor, London Breed, discussed changing planning laws to make it easier to convert commercial buildings into housing—one estimate suggests the city’s empty offices could hold 11,000 new homes.

The problem for many cities, particularly in the US, is the buildings that populate their downtowns. Some are too big to easily transform into apartments with access to natural light. In New York, Freemark says, it’s proven much simpler to convert the smaller-footprint 1920s and 1930s buildings of downtown than the square-sided Midtown behemoths of the 1960s and 1970s.

Like bamboo or Japanese knotweed, block-wide skyscrapers can also stifle life at ground level. They have a “deadening” effect, Freemark says. “The forms of the buildings are incredibly hostile to pedestrians,” Freemark says. “The park space is minimal. Roadways are horrible and extremely car-focused. All those things have to be thrown out if you want to create a neighborhood.”

In that sense, the city by the Bay, far from being a cautionary tale, might actually be a model for the future. “San Francisco, in many ways, does it better than almost any other US city,” Freemark says. It has vibrant, walkable neighborhoods that people want to live in, even now. Like hundreds of other cities around the world, it just needs to figure out how to turn downtown into one of them.

You can’t dig through the WIRED archives for stories about cities without reference to the iconic piece “Disneyland With the Death Penalty,” by William Gibson, which sent the sci-fi visionary to Singapore’s “clean dystopia.” Since the article was published in 1993, these kinds of cities have proliferated, particularly in the Middle East. But they also sprout in pockets throughout major capitals—private islands in central London and New York with their own security guards, pseudo-public spaces that you can walk through but not loiter in.

The sensation of trying to connect psychically with the old Singapore is rather painful, as though Disneyland’s New Orleans Square had been erected on the site of the actual French Quarter, obliterating it in the process but leaving in its place a glassy simulacrum. The facades of the remaining Victorian shop-houses recall Covent Garden on some impossibly bright London day. I took several solitary, jet-lagged walks at dawn, when a city’s ghosts tend to be most visible, but there was very little to be seen of previous realities: Joss stick smoldering in an old brass holder on the white-painted column of a shop-house; a mirror positioned above the door of a supplier of electrical goods, set to snare and deflect the evil that travels in a straight line; a rusty trishaw, chained to a freshly painted iron railing. The physical past, here, has almost entirely vanished.

Greg asks, “In the first season of Better Call Saul, a small case of nursing home overbilling ballooned into a multistate corporate fraud. The overwhelming amount of research required meant that Saul’s baby had to be parted out to two huge firms.

It occurred to me that this would be a perfect job for AI, at least for the first pass of sifting through reams of cases for relevant precedents. Yet in all the articles I’ve read, the focus is limited to AI’s impact on creatives, specifically writers and artists.

Law and, as I think of it, medicine, would seem to be perfect candidates for this kind of brute-force research. Thoughts?”

Hi Greg. Cynically, one reason that a lot of articles have focused on AI’s impact on writers and artists might be that they’re the ones creating the articles—to some journalists, AI has become a threat to their livelihood

But the two forms of AI that have exploded in popularity in the past six months or so—large language models like ChatGPT and image-generation apps like Midjourney—are generative AIs, tailored towards creating text and creating images. In many cases they’re doing a scarily good job at it. 

The problem with trying to use something like ChatGPT, or even Google Bard or similar, for law or medicine is that it doesn’t actually understand legal or medical concepts—it’s based on a probabilistic approach to language. That will serve you the most likely answer, not necessarily the correct answer. It’d be a brave attorney who relied on the analysis of a chatbot in court without manually checking it.

For now, I think AI will be most useful to lawyers for more procedural work. In March, researchers showed that GPT-4 can reach a passing mark on the bar exam, and the number of legal AI startups is growing. One called Harvey is building custom large language models for law firms, and Allen & Overy has already said its lawyers will be able to use the technology to automate tasks like document drafting and research.

You can multiply that by a thousand for medicine, where (accelerated by the pandemic) AI is being used for everything from sifting through research papers to designing new drugs to analyzing mammograms. Ultimately, AI is such a broad term, encompassing so many different types of technology, that for any industry you can think of, there’s probably someone trying to disrupt it with something that they’re calling artificial intelligence.

You can submit questions to mail@wired.com. Write ASK LEVY in the subject line.

A free 4K TV, with a smaller screen underneath that constantly shows you advertisingeven when you’re not watching TV. 

Google’s Bard chatbot isn’t available to the 447 million people in the EU—but it has been released on Bouvet Island, a Norwegian territory home to 50,000 penguins

Global suicide rates have fallen by a third in the past few decades, but the US is bucking the trends. You’ll never guess why.

AI language models have been reading (and plundering) erotic fan fiction.

Don’t miss future subscriber-only editions of this column. Subscribe to WIRED (50% off for Plaintext readers) today.