A society of 25 artificial intelligence (AI) robots have been observed waking up, cooking breakfast, commuting to work, going to the bar for lunch with friends and even throwing a party, according to six researchers who created a city ​​for them.

Google and Stanford University researchers wrote in an April 7 article titled “Generative Agents: Interactive Simulacra of Human Behavior” that they had built a virtual city populated by ChatGPT-trained “generative agents.”

The aim of the study – which has not yet been peer-reviewed – was to create a small interactive society of AI robots inspired by life simulation games such as The Sims.

Agents could make a wide range of inferences about themselves, other agents, and their town of “Smallville” by synthesizing new information, storing it in memory, and then behaving in a way that reflects that knowledge. .

For example, agents can turn off their stove when they see their breakfast is burning, coordinate plans, and even engage in seemingly meaningful conversations with other agents.

The results led the researchers to conclude that the generative agents produce “believable” human behaviors:

“By fusing large language models with interactive computational agents, this work introduces architectural and interaction models to enable believable simulations of human behavior.”

An example shared in the study explained that AI agent “Isabella Rodriguez” invited nine other agents to a Valentine’s Day party at the town cafe.

Details of the party were passed on to four other people, including “Abigail”, who then expressed their excitement for the upcoming event with Isabella.

In another example showing the “day in the life” of an AI agent, “John Lin” woke up at 7 a.m., brushed his teeth, took a shower, took his breakfast and checked the news at his living room dining table.

Before John’s son, Eddy, left for school, John asked what he was going to be working on for the day. Eddy replied and John commented on it before sharing the news with his “wife”, Mei.

However, not all went well in the experiment.

As each AI bot’s memory grew with each interaction, the most relevant information might not be retrieved, and as a result “some agents chose less typical locations for their actions.”

Related: Elon Musk and technical leaders call for a pause in AI development

For example, when officers had to decide where to have lunch, many chose the cafe in town first. However, the researchers said:

“When some officers learned of a bar nearby, they opted to go there for lunch instead, even though the bar was intended as a hangout later in the day, unless the town has spontaneously developed a habit of drinking in the afternoon.”

In another example, some AI agents walked into stores in Smallville that were closed, while college students met others in the dorm bathroom because they thought it might be busy. by more than one age.

The researchers said they will soon develop the “expressiveness” and “performance” of AI robots using the more advanced GPT-4, the latest iteration of ChatGPT, which passed high school and state law exams. States in the 90th percentile.

Magazine: NFT Creator, Emily Xie: Creating “Organic” Generative Art from Robotic Algorithms