Can ChatGPT Help Save Lives?

Supplied with well-crafted prompts, the technology proved adept at locating people stranded in a natural disaster.

illustration of first responders.

People caught in a natural disaster often plead for help on social media when 911 systems become overloaded. Yet first responders don’t have the resources to monitor social media feeds at such a time, to follow the various hashtags and decide which posts are most urgent. 

This led a UB-led research team to wonder: Could ChatGPT be trained to recognize location information in tweets, and lead first responders to those in trouble? The answer, according to their study, is yes.

Fusing ‘geoknowledge’ into ChatGPT

Imagine a tweet with an urgent message: A family of four needs rescuing at 1280 Grant St., Cypress, Texas.

A typical model, such as a named entity recognition (NER) tool, would recognize the listed address as three separate entities: Grant Street, Cypress and Texas. If this data was used to geolocate, the model would send first responders not to 1280 Grant St., but into the middle of Grant Street, or even the geographical center of Texas.

According to Yingjie Hu, associate professor of geography at UB and lead author of the study, NER tools can be trained to recognize complete location descriptions, but it would be a labor-intensive and time-consuming process.

“First responders have a lot of knowledge about the way locations are described in their local area, whether it be the name of a restaurant or a popular intersection,” said Hu. “So we asked ourselves: How can we quickly and efficiently infuse this geoknowledge into a machine-learning model?”

The answer was OpenAI’s Generative Pretrained Transformers, or GPT, large language models already trained from billions of webpages and able to generate human-like responses. Through simple conversation and the right prompts, Hu’s team thought GPT could quickly learn to accurately interpret location data from social media posts.

A test case: Hurricane Harvey

To train the system, researchers provided GPT with 22 real tweets from Hurricane Harvey victims, noting which words in the post described a location and what kind of location they were describing.

They then tested the geoknowledge-guided GPT on another 978 Hurricane Harvey tweets, and asked it to extract the location words and guess the location category by itself.

The results? The geoknowledge-guided GPT models were 76% better at recognizing location descriptions than non-guided GPT models, and 40% better than NER tools.

The team hopes their work will lead to AI systems that automatically process social media data for emergency services, helping first responders reach victims more quickly and ultimately saving more lives. 

“ChatGPT and other large language models have drawn controversy for their potential negative uses, whether it be academic fraud or eliminating jobs,” said Hu. “So it is exciting to instead harness their powers for social good.”