"911, please name your emergency." In the United States, when this familiar voice appears, it means that help is coming. But if you were told that artificial intelligence may have stood behind that voice, how would you feel?
That's right, a quiet AI revolution is underway in the United States. Although it has not yet taken over the call directly, AI has quietly become an important assistant in emergency rescue. In the future, it will intervene in more ways and assist our ambulance system.
There are about 240 million 911 calls in the United States each year. While not every phone call is a true emergency, it still puts a lot of pressure on the operator. At the same time, the police system is facing a serious shortage of manpower, and 911 wiring positions are "talent shortage", and many experienced staff are lost.
New operators have little work experience, and language barriers make it more difficult for non-native English speakers to seek help. A variety of factors contribute to errors and delays in response, which will only worsen if not addressed soon.
In this context, artificial intelligence has been given a new task - to assist in emergency rescue. Unlike the imagined AI that will replace human work, AI here plays a more auxiliary role, reducing the burden on operators by automating simple repetitive tasks, while using computing power to integrate and analyze large amounts of data.
At present, the main role of AI is to help filter non-emergency calls. Cities like Los Angeles and Austin are already using AI to process reports from non-emergency helplines, such as repairing stagnant puddles and rescuing cats from trees. This frees up more time for operators to deal with truly urgent situations.
At the same time, speech-to-text technology is also widely used in 911 systems. This is similar to the instant captioning technology we use on Zoom to convert speech into transcripts. This can not only record the content of the call, but also automatically mark out the key information needed for rescue, such as the location of the incident, the nature of the incident, etc., reducing the operator's workload of taking notes.
But that's just the basics. To be truly effective, AI needs to gain an understanding of the "context" and accumulate a lot of "experience." So far, these are concepts that have been difficult for AI to grasp.
However, the recent emergence of large-scale language models such as ChatGPT has obtained a kind of "experience" prototype by memorizing massive amounts of information and establishing semantic associations. This opens the door to new applications of AI in emergency rescue:
- Using multilingual speech models, AI can detect emergency calls in different languages, greatly reducing language barriers;
- AI can analyze past rescue calls and situations, provide operators with suggestions for dealing with new situations, and make up for inexperience;
- AI can create response workflows synchronously to ensure that police, medical personnel, etc. are notified;
- AI can identify system vulnerabilities through quality assessment and help management develop measures such as personnel deployment.
When AI truly becomes a partner in the 911 system, it will have the power to save lives. It may not be a sensation then, but it will quietly play the role of a hero.
Of course, testing remains a huge challenge for AI. Every emergency call is unique, making it difficult to fully assess AI readiness. It must gain experience in non-emergency situations to handle real life and death calls. But the process has already begun.
The digital upgrade of rescue systems is the trend of the times, and human beings also need to keep pace with the times. The rational and prudent use of AI will be an important help for us to respond to various emergencies.
Maybe soon, when you call 911, you'll get direct help from AI. It will hear the situation, locate the location, and immediately deploy the quickest rescue measures. At that point, you'll find that cold machines will also be humanity's most reliable friend.