The T-Shirt Invisibility Cloak
0 up · 0 down · 0 ratings
Promos
Visit brilliant.org to get started learning STEM for free, and the first 200 people will get 20% off their annual premium subscription. Learn how adversarial patterns in simple clothing can completely fool AI-driven surveillance cameras. Leave a reply with your requests for future episodes. ► GET MERCH: lttstore.com ► LTX 2023 TICKETS AVAILABLE NOW: lmg.gg ► GET EXCLUSIVE CONTENT ON FLOATPLANE: lmg.gg ► SPONSORS, AFFILIATES, AND PARTNERS: lmg.gg FOLLOW US ELSEWHERE --------------------------------------------------- Twitter: twitter.com Facebook: @LinusTech Instagram: @linustech TikTok: @linustech Twitch: twitch.tv
The video introduces the concept that you can defeat high tech surveillance using a simple, low tech tool: clothing that contains adversarial patterns. It explains that certain AI facial recognition systems, like YOLO V2, can be reverse engineered to misinterpret images, making a person appear as not a person or even as a different object. The host describes how these adversarial patterns often resemble abstract, bright art and how they can render a wearer effectively invisible to AI while still being recognizable to other humans. The discussion covers that these patterns are not foolproof and can be defeated by more advanced algorithms released in the future, meaning the technique is not a guaranteed shield. The video also presents practical alternatives to fool AI cameras, such as clothing that imitates objects or masks with lens-like effects, which can obscure identity while remaining socially legible. It broadens the scope by noting that many surveillance systems around the world are not just academic curiosities but widespread tools for traffic management, crowd security, and law enforcement in smart cities. Finally, it cautions viewers about the limitations and potential risks of attempting to evade surveillance, and it encourages responsible engagement with future episodes and topics.
Topics · science and technology · privacy and surveillance · computer vision · fashion technology · surveillance technology
Questions answered
- What is YOLO V2 and why is it relevant to this topic?
- YOLO V2 is a real time object detection algorithm used for recognizing humans and objects in images, which researchers have shown can be influenced by adversarial patterns to misclassify or fail to detect a person.
- Do adversarial clothing patterns guarantee invisibility to cameras?
- No, they do not guarantee invisibility. The effectiveness depends on the algorithm, lighting, camera quality, and the specific adversarial pattern, and more sophisticated systems may still recognize a person.