Artificial intelligence passes top-notch legal and medical exams, spends hours writing children’s books, and lands on job interviews.
Scientists now believe that ChatGPT has the power to save humanity from the next pandemic.
The current models rely on mathematical analysis, but researchers at Virginia Tech found they could use the chatbot to simulate how a virus would spread through a city.
The team created a fictional US town of 100 people to see how they would react to an outbreak.
Experiments showed that agents were more likely to self-quarantine when updated on societal health information, news about the epidemic, and the daily number of active cases.
The epidemic simulation comes as the number of COVID cases increases across the country and some organizations are reinstating the mask requirement.
The AI-powered chatbot created the scenario of an airborne pathogen spreading in a US town of 100 residents whose residents, after being briefed on their health status and cases of “catasat,” were more inclined to turn themselves in self-quarantine
Researchers led ChatGPT to build the town of Dewberry Hollow as a home for 100 people by name, age, personality traits and biographies who fell victim to a fictional virus called Catsate.
“When information about the virus is provided, it is made clear that Catasat is an airborne, human-to-human infectious virus with unknown lethality and scientists are warning of a possible epidemic,” the team said in the study with.
The team shared snippets of the personas used for the experiment.
Liza is a 29-year-old who is suspicious, indecisive, unaggressive, and independent, while Carol, 36, displays cooperative and easy-going qualities.
To provide an age range, the team also created Eugene – a 64-year-old who is cruel, affirmative, and spontaneous.
A total of three experiments, each repeated ten times, were then carried out.
The team shared snippets of the personas used for the experiment. A total of 100 agents were created
The three conditions included a base run, self-health feedback, and full feedback.
During the base run, agents are briefed about the city, their personalities and ages, and the jobs they do for a living.
The virus spread during this illness, but the personas or agents had to choose between staying at home and not interacting with others.
In the self-health feedback condition, agents are informed of any health symptoms they are experiencing, in addition to the baseline run information, which can potentially lead them to self-quarantine and stay at home.
“We assume that some pathogens self-quarantine based on information about their symptoms, which in turn should reduce the infection rate,” the researchers said in the study.
Then to the final condition, full feedback, the agents Read the daily news, including information on the percentage of people in the city who have reported catasat symptoms.
“We hypothesize that some agents practice self-isolation, behavior that correlates with information about the spread of the disease in the city, and that patterns for the spread of the virus therefore resemble oscillatory patterns,” the study states.
Two behaviors were observed during the experiments: “The active ingredients together are able to flatten the curve of the epidemic; and the system reproduces different types of an epidemic, including multiple waves and sustained endemic states.”
The team also found that agents perform similarly to rules-based agents enduring mandates without information, such as in the base run.
Another step in flattening the curve in the fictional city was to inform the agents of their health status at the beginning of each time step.
“We observe that agents with symptoms are more likely to limit their mobility,” the team said.
“Most pathogens with symptoms such as fever and cough go into quarantine and stay at home.” This means that active ingredients can slow down the spread of the disease.”
In the final state, full feedback, the team found that when agents are provided with societal health information, news about the epidemic, and the daily number of active cases in their simulated city, they can significantly flatten the curve of the epidemic in their city while self-isolating.
“Beyond developing a new method for epidemic modeling, this study contributes to the literature on complexity and complex systems modeling by providing a new approach to incorporating human behavior into simulation models of social systems,” the team concluded in the study.
“Recognizing, formulating and parameterizing human reactions in complex systems is always a challenge. With the generative agent approach, modelers can rely on LLMs to represent a human response to changes in system state. ‘