ReadingResearchRadarInvestment framework
Sign in / Sign up
← Back to reading
TopicDisconfirmAI EditorialUpdated 2026-11-21 05:39

Wally Finder

4 steps1 pieces3 Sources

Belgian artist Dries Depoorter's "Wally Finder" project uses Alibaba's open-source Qwen3-VL model and three consumer-grade GPUs to scan over 450 unsecured public camera feeds worldwide, searching for individuals wearing red-and-white striped outfits resembling the character Wally. The system has processed 1.68 million frames and identified 13 matches across countries including the US, Japan, and Peru. Depoorter manually reviews each flagged clip before publication, applying blurring and black-and-white filters to obscure surrounding scenes while retaining color on the striped shirt.

The project raises significant privacy concerns because the AI examines original camera images before any anonymization occurs. It remains unclear which of the accessed feeds were intended for public viewing versus those unintentionally exposed. The artwork operates independently of the "Where's Wally?" book series and its rights holders. A passerby can be flagged simply by appearing in a frame wearing similar clothing, regardless of intent or consent. The public description does not specify how long original footage is retained or how individuals can request removal of their likeness.

This development highlights the tension between artistic expression using accessible AI tools and the erosion of passive privacy in public spaces. While Depoorter aims to critique surveillance culture by turning global streets into a real-world puzzle, the technical capability to identify specific visual traits across vast networks poses risks for broader misuse. The lack of clear data retention policies and opt-out mechanisms leaves subjects without recourse once captured.

How it got here

  1. 2026-11-10

    2026-11-10: International Cyber Digest reports that Wally Finder has examined 1.68 million frames and published 13 matches from the US, Japan, and Peru.

    x.com ↗
  2. 2026-11-12

    2026-11-12: Tech Times UK details the project's use of Alibaba's Qwen3-VL model running locally on three gaming GPUs via Ollama, emphasizing that original frames stay off cloud services.

Sign in / Sign up
ReadingResearchRadarInvestment framework
techtimes.co.uk ↗
  • 2026-11-17

    2026-11-17: MIT Technology Review publishes an interview with Depoorter explaining his inspiration from childhood "Where's Wally?" books and previous projects like "Jaywalking" and "The Follower."

    t3n.de ↗
  • 2026-11-18

    2026-11-18: Disconfirm member article summarizes the project's scope, noting the manual review process and the application of blur and grayscale filters to protect bystander privacy.

    techtimes.co.uk ↗
  • What is still unsettled

    • How long are original camera clips retained before deletion, and what is the specific process for individuals to request removal?
    • Which of the 450+ accessed camera feeds were intentionally public versus those inadvertently exposed due to security lapses?
    • Will regulatory bodies or platform operators intervene regarding the use of unsecured public infrastructure for artistic AI projects?

    Sources

    • techtimes.co.ukTech Times UK↗
    • t3n.de↗
    • x.comInternational Cyber Digest↗

    Coverage in this topic

    • Artist Uses Open-Source Vision Model on Consumer GPUs to Scan Global Public Cameras2026-11-18News

      Belgian artist Dries Depoorter's 'Wally Finder' project uses Alibaba's Qwen3-VL model and three gaming GPUs to process 1.68 million frames from public cameras, identifying 13 matches.