• Researchers demonstrate that misleading text in the real-world environment can hijack the decision-making of embodied AI systems without hacking their software.
  • Self-driving cars, autonomous robots and drones, and other AI systems that use cameras may be vulnerable to these attacks.
  • The study presents the first academic exploration of environmental indirect prompt injection attacks against embodied AI systems.

Photos

all 12 comments

sorted by: hot top controversial new old
[–] 110 points 9 months ago (5 children)

An article about the dangers of AI starts off with an AI generated comic about the dangers of AI.

  • source
  • hideshow 5 child comments
  • [–] 39 points 9 months ago (1 child)

    This whole article reminds me so much of the rogue AI sign found in Portal 2

  • source
  • hideshow 1 child comment
  • [–] 13 points 9 months ago

    One time I saw a 30mph sign spray painted to say 88mph speed limit. Good thing it was before self driving cars of that would have been crazy.

  • source
  • [–] 5 points 9 months ago

    That is actually hilarious 😂

  • source
  • [–] 1 point 9 months ago

    Misleading text in the physical world can hijack AI-enabled robots, cybersecurity study shows

    Eh, yes? Hasn't that been like that since day 1?

  • source
  • [–] 1 point 9 months ago

    that use cameras

    Make it embossed letters for all other types of sensors.

  • source