AI Mistook Doritos for a Gun, Cops Mistook a Child for a Threat
An AI tool, Omnilert, mistook a bag of Doritos for a firearm, leading Baltimore school police to "swarm and handcuff" a student. This isn't just an AI glitch; it's the inevitable outcome of outsourcing policing to uncritical technology. These systems are sold as objective, yet they are trained on data reflecting existing biases, often leading to disproportionate scrutiny of certain communities.
The 'threat detection' industry, a multi-billion-dollar market, profits regardless of whether the 'threat' is a weapon or a chip crinkle. Instead of questioning the fundamental principle of policing children with algorithms, the conversation pivots to 'improving AI.' Meanwhile, the surveillance apparatus expands into schools, funded by tax dollars that could instead address root causes of
violence, like poverty and underfunded education. One might wonder, how many innocent children must be traumatized before we question the wisdom of letting silicon chips declare a threat where none exists?