Hi, welcome to your Weekend!
We all have our spot: The one cafe within a 5-minute walk (or a 10-minute drive, if you work in L.A.), where you can grab a quick meeting, sidle up to an acquaintance or eavesdrop on a stranger. It’s the kind of place where you feel like you’re being productive, even when you’re really procrastinating.
My favorite coffee hang in San Francisco is Noe Cafe, which is roughly midway between my house and my office, and squarely located in the techie-rich corridor of Noe Valley. The outdoor seating? Plentiful and shaded. Traffic noise? Muted. Coffee and pastries? Delicious. And I always see someone I know...or would like to know.
For this week’s cover, Annie spent the past several weeks querying venture capitalists on both coasts for their own personal Noes. (Some even recommended Noe Cafe as well—I’ll see you there, Martina Lauchengo and Megan Quinn!) The resulting list of 25 cafes is a caffeine lover’s dream—albeit a nightmare for a few privacy–coveting VCs.
A few of those investors groused to us about this project: Turns out, not everyone wants their secret corner cafe overrun by competitors—or by thirsty founders desperate to fundraise. We can sympathize with that. But isn’t it nice to know that so many of these local gathering spots, once abandoned, are thriving again?
The VC cafe is dead. Long live the VC cafe.
Venture Capital’s 25 Favorite Cafes
It’s been a Silicon Valley truism for years: If you want to “accidentally” bump into a venture capitalist, camp out at their favorite coffee shop. To assemble a comprehensive map of the current cafe landscape, Annie asked over 40 VCs for their go-to java spots in the Bay Area, New York and LA. Run into them at your own risk.
An AI Researcher Tries to Build Good AI While Burning Down the Bad
Margaret Mitchell was fired by Google after publishing a critical research paper about its AI efforts. Now she’s the chief AI ethicist at one of its startup rivals, Hugging Face.
Silicon Valley’s Realtors, Like Its Bankers, Are Having a Tough Month
The regional bank meltdown has created a big problem for the Bay Area’s real estate agents: Their tech clients just lost access to sweetheart mortgage deals. Pair that with the downturn in the housing market and mass layoffs within big tech and, as Zara Stone reports, it’s been rough going for realtors.
The Skin-Tech Devices Helping Execs Beautify in a Hurry
There are now a dizzying number of gadgets promising to bring the dermatologist’s office home, zapping, toning and lifting skin to taut perfection. Veteran beauty editor Aja Mangum asked founders and CEOs from the tech and beauty industries to divulge their favorite new devices.
Watching: Requiem for the Roys
The P.J.s—private jets, duh—touch down in T-minus 35 hours and counting. Ahead of the final season of “Succession” premiering tomorrow night on HBO, showrunner Jesse Armstrong has promised a “muscular season to exhaust all of our reserves of interest” in the Roy clan. By early accounts, Armstrong delivers on his pledge—with the BBC describing Season 4 as “jaw-dropping,” while Rolling Stone deems it a “guns blazing” conclusion. (Or, as the Times of London puts it, a “last serving of delicious snake pie.”) I asked The Information’s founder and CEO, Jessica Lessin, whether she thought “Succession” accurately reflected the lifestyles of media moguls. “Seems pretty accurate to me,” she emailed. “I watch the show rabidly both for what TO do and what NOT to do in running a family media business.” And here I am just watching it for the memes. —Abe
Generating: ChatGPT-4’s misinformation problem
When OpenAI’s newest large language model launched last week, the company issued a series of reassurances. It said ChatGPT-4 was “significantly” less likely to produce falsehoods than the previous version, ChatGPT-3, and 82% less likely to respond to requests for sensitive content, like methods of self harm. But a study by misinformation watchdog NewsGuard showed otherwise. The site asked both versions of ChatGPT to generate propaganda, feeding the models prompts about the Sandy Hook school shooting, quack cancer treatments and faked alien sightings. ChatGPT-3 was more likely to refuse to answer these prompts altogether; when both models did answer, ChatGPT-4’s responses were longer and more persuasive—which the researchers believe could make them more pernicious. OpenAI insisted that the path to a safer model will be through trial and error. The question is how many scams and how much misinformation will emerge in the meantime. —Margaux
Experiencing: A gallery of AI delights
All that AI bed-wetting got you down? Treat yourself to a big gulp of creative optimism at San Francisco’s Misalignment Museum. The new pop-up art gallery in the Mission District (free to visit through May 1) imagines a future in which humans and artificial intelligence might just get along. On a visit this week, Weekend editor Jon and I listened to part of an “infinite conversation” between two deepfaked geniuses: philosopher Slavoj Žižek and filmmaker Werner Herzog. We watched a group of Spam Bots (literally, cans of spam attached to keyboards) peck out an original novel. And Jon got mauled by an oddly aggressive Broomba (a Roomba device carting around a broom). The Misalignment Museum’s curator, Audrey Tang, told us she got interested in AI while working on autonomous vehicles at Cruise. She still believes that AI could save myriad lives; on the other hand, who is making sure AI models align with human values? And which human values? Before exiting her museum, we begged for mercy in the Church of GPT, a voice-activated installation powered by AI. Sadly there was no answer. —Arielle
Makes You Think
Until next Weekend, thanks for reading.
Weekend Editor, The Information