AI is all around us, but that doesn’t mean we can’t push back against it, fight it, and subvert it. Sometimes the best way to intervene with AI, to rebel against AI, is to get creative. We’ve found and outlined a few different ‘interventions’ you can take at home to trick and trip up. You can add noise to the machine by altering data sets, use fashion to obfuscate computer vision systems, document AI in your daily life, and build new alternatives to algorithmic oppressive systems.
Creative Interventions
These creative interventions are inspired by anti-surveillance rebellion and the works of Octavia Butler. Mutale Nkonde mentioned using avant garde makeup to trick or confuse facial recognition systems, a tactic uncovered and made famous by researcher and artist Adam Harvey in his CV Dazzle project. Technologist and artist Joselyn McDonald goes a step further by creating intricate flower face crowns to shield one’s face from facial recognition. Her suggestion is that you turn on your phone, open up INstagram and use a face filter to test your flower shield or CV dazzle. Keep adjusting until the face filter doesn’t recognize your face.
Here are some examples we love:
- CV Dazzle by Adam Harvey
- Mother Protect Me by Joselyn McDonald
- Another creative intervention is this sweatshirt that tricks license plates readers and causes them to malfunction. Created by the punnily named Adversarial Fashion
- Journalist Paris Martineau shared her sweatshirt she purchased
- Print that tricks to counter facial recognition and you can put it on your mask
Creating Noise in the Machine
Facebook seems to know everything about us, from what we like, where we live, and who we are friends with. From this data they gather about us, Facebook is able to use targeted advertising systems and create recommendation algorithms and filter bubbles to serve people content, sometimes to disastrous and dangerous results. But there are ways to trip up Facebook’s algorithms and we have a few suggestions below. You can use these same suggestions on any big tech social network like Instagram, TikTok, Twitter and others.
- Trick 1: Unfollow everyone on facebook (but you’ll still be friends). What do you see? What does your feed look like? Is it it empty, strange, bizarre?
- Trick 2: Change some key information about yourself that is wrong: maybe it’s your location, age, relationship status, gender, birthday or favorite movies. What kinds of ads or content are you seeing now?
- Create a brand new profile, and follow a few well known figures you would not normally follow. What do you see here? What content is suggested? Reflect on how or why the algorithm is suggesting this.
Tricking Computer Vision and AI
Change your images! Using the new software “Fawkes”, upload an image to Fawkes, which adds in ‘digital noise’ that’s invisible to the human eye to that image. What Fawkes does is it tricks Facebook into thinking your image is another person. It doesn’t completely cloak you but like wearing a mask, it makes you look different to the computer vision systems on Facebook. Now, upload the image to Facebook and see if the auto tagging feature (that uses facial recognition) works!
Documenting Your Day to Day
AI is all around us, in almost everything we use. Try this activity inspired by the People’s Guide to AI, to better ‘see’ the AI in your daily life. Take a piece of paper, and throughout your regular day, write down when you think you use AI. For example, when you use a search engine or even a search function on a social network, write that down. Try opening up Netflix or Spotify. What are the recommendations you see?
Learn More!
There’s still so much to talk about and learn in the AI space. We’ve selected a few of our favorite articles and projects on confronting AI, creating better AI, and creating better systems online. A way to create a better world is to decolonize technology, start crafting exits from surveillance capitalism, understanding when technology won’t fix a problem, and creating better responsible systems that center marginalized communities and their needs.
