New Orleans is testing just how much Americans are willing to trust artificial intelligence when the stakes are considerably higher than writing an email or ordering dinner, according to the New York Post.
The city has begun allowing an AI system to field certain 911 calls, apparently making it the first major U.S. municipality to put a bot directly between emergency callers and human operators.
The idea is largely about volume. New Orleans' emergency communications center receives more than 1,000 calls on a typical day, and officials believe automation can clear some of the clutter while dispatchers concentrate on genuine emergencies.
The technology isn't supposed to take over the entire 911 operation. Instead, it can deal with situations that appear redundant or relatively routine when human operators are tied up. A flood of calls about the same already-reported car accident, for example, could potentially be filtered through the automated system rather than occupying multiple dispatchers.
The Post writes that city officials insist humans remain at the center of the operation. But skeptics say emergency calls are precisely the kind of messy, unpredictable interactions computers may struggle to interpret.
People dialing 911 aren't reading from scripts. They may be terrified, whispering, intoxicated, injured, confused or speaking through heavy background noise. In some situations, callers intentionally conceal what's happening because another person is nearby.
One online critic who identified himself as a former 911 dispatcher called the experiment “the worst idea you could possibly think of,” arguing that experienced operators routinely listen for clues beyond the caller's literal words.
That concern becomes more significant given the continuing reliability problems surrounding conversational AI. Voice models can confidently produce incorrect information, while cybersecurity researchers have shown that specially constructed audio can sometimes manipulate AI systems into behaving in unintended ways.
New Orleans isn't alone in bringing algorithms into emergency services. Atlanta uses AI technology to assist operators in locating callers, while Seattle has used automated tools to help sort medical calls by urgency. The distinction is that New Orleans appears to be going further by allowing AI to actually interact with some people dialing 911.
Officials believe safeguards, human supervision and strict limits on when the system is activated can make the technology useful without handing emergency response over to a machine.
It's an ambitious experiment, although perhaps an uncomfortable one: the next time somebody calls 911 in New Orleans, the first voice answering may not belong to a person at all.