Surveillamancy

For this week’s Electronic Rituals, Oracles, and Fortune Telling meditation we are to:

Invent an “-omancy,” or a form of divination/prophecy based on observing and interpreting natural events. Your reading of “natural” should make some reference to digital/electronic/computational media. (What counts as a “natural event” on the Internet? What’s the electronic equivalent of phrenology, from both a physical computing perspective and a data analysis perspective? Does it count as “interpretation” if it’s being performed by a computer program?) I’m especially interested in responses that take the form of purposefully inaccurate data analysis.

I’ve created surveillamancy, a form of prophecy that is generated off of what is seen when surveilling a camera somewhere in the world. It is accessed at https://surveil.netlify.com/.

In it the viewer selects which camera tickles their fancy:

Then the viewer waits as the camera is located.

Finally, the user sees a picture captured from the camera and an interpretation of meaning derived from the picture’s contents. Artificial intelligence is used to determine what is in the photo.

What makes it an ‘omancy?’

As we learned in class, the general characteristic of an ‘omancy’ or prophecy is if x then y. The prediction/prophecy should be based on something discreet, even if the connection between them is nonsensical.

In the case of surveillamancy, the x in if x is both the viewer’s choice in which type of camera, and the contents that exist in the photo. AI’s interpretation of the contents and the connection with a meaning are the y.

How does it work?

The code can be seen at github.com/oveddan/surveillamancy. It is written in react on the client side and node.js on the server, and is hosted on Netlify and Heroku.

To find a camera of the specified type, https://www.insecam.org/ is scraped by going to https://www.insecam.org/en/bytag/{cameraType}/ and a random camera that appears is selected. Puppeteer is used to do the scraping. The camera is then navigated to, and a screenshot is captured of the video feed. The code for this can be seen in server/insecam_scraper.js.

To derive meaning from the contents of the photo, clarifai’s image classification api is used to classify what’s seen in the photos.

To derive meanings from this classifications, Gustav Hindman Miller’s Ten Thousand Dreams Interpreted was scraped for dream meanings. The scraper can be seen at server/scrape_dreams.js

Then, at runtime, the dream interpretations for the classifications from clarifai are looked up in these dream interpreations. All instances of dream are replaced with see:

function changeToSoundLikeVision(dreamMeaning) {
  return dreamMeaning
    .replace('To dream of', 'To see')
    .replace('to dream of', 'to see')
    .replace('To dream that', 'To see that')
    .replace('to dream that', 'to see that')
    .replace('dreamer', 'voyeur')
    .replace('dreaming', 'voyeuring')
    .replace('dream of', 'see')
    .replace(' in a dream', '')
    .replace(' in your dream', '')
}

The first 3 interpretations are displayed to the viewer.