AI Summit_Sept. 13 2024

Hypothetical 4 (“ Security AI ”):

• A Security AI System uses cameras and sensors comparing inputs with object signatures to identify weapons and potential security risks. • It was trained on a police database of images of deidentified people carrying different weapons and objects and improves with usage as it collects more data. • When it identifies a match, security personnel are sent images of the suspected individual for further human intervention and assessment. • The Security AI System is deployed at the entrance of a gov’t building. • It flagged Joanne, a disabled woman holding a cane, as someone who may be carrying a weapon, and she was stopped at the entrance.

16

Practical Takeaways for Resolving Complex Dilemmas

AI Roundtable Page 669