A screenshot of City Describer, a Twitter bot that runs off Microsoft's "custom vision" AI technology. Geoff Boeing/City Describer

L.A. looks like "a Cactus." Is that plain wrong? Or is our hubris keeping us from recognizing the profound meaning behind these captions?

Los Angeles. City of Angels. La La Land. Where dreams of fame materialize and evaporate like heat mirages on hot asphalt. This city has served as the muse of surrealist directors, retro-glam pop divas, and visionary rappers. To some, its sprawling sepia landscape evokes yearning and melancholia; for others it is marked by grit and resilience.

Or, you know, it could also be described thusly:

That’s how the Twitter bot aptly named City Describer  *sees* this Reddit-sourced image of L.A. The bot is designed by Geoff Boeing, an urban data wonk at UC Berkeley (whose research and urban data tools we’ve written about in the past) using Microsoft’s open-source “custom vision” AI technology. And man, the bot delivers—although not always in the way you’d think.

Its descriptions of cityscapes around the world range from accurate to delightfully absurd. For one, the bot can see ghosts:

At times, these visual interpretations sound like cheeky, Clickhole-esque disses of city porn. If you consider yourself a connoisseur of urban form, they might annoy you:

Or, you know, if you’re religious. This bot DGAF about your sensibilities:

But by LOLing at these tweets, are we being facetious? Are these misreadings really misreadings? Or is even fledgling AI already making deductions about the human race so profound that our hubris prevents us from recognizing it? (#Neverforget when that Microsoft Twitter bot learned to be super racist, like, within a day.)

Really think about City Describer’s caption for L.A., for a second. The more enlightened among us will perhaps be able to see it for what it really is: an astute distillation for the city’s many complexities—a version of answer “42” to the question of Life, the Universe, and Everything.

>

About the Author

Most Popular

  1. Equity

    How Poor Americans Get Exploited by Their Landlords

    American landlords derive more profit from renters in low-income neighborhoods, researchers Matthew Desmond and Nathan Wilmers find.

  2. An illustration of a private train.
    Transportation

    Let’s Buy a Train

    If you dream of roaming the U.S. in a your own personal train car, you still can. But Amtrak cuts have railcar owners wondering if their days are numbered.

  3. A photo of the interior of a WeWork co-working office.
    Design

    WeWork Wants to Build the ‘Future of Cities.’ What Does That Mean?

    The co-working startup is hatching plans to deploy data to reimagine urban problems. In the past, it has profiled neighborhoods based on class indicators.

  4. A photo of a new subdivision of high-end suburban homes in Highland, Maryland.
    Equity

    Unpacking the Power of Privileged Neighborhoods

    A new study shows that growing up in an affluent community brings “compounding privileges” and higher educational attainment—especially for white residents.

  5. A photo of San Antonio's Latino High Line
    Equity

    A 'Latino High Line' Promises Change for San Antonio

    The San Pedro Creek Culture Park stands to be a transformative project for nearby neighborhoods. To fight displacement, the city is creating a risk mitigation fund.