A recent study showed that AI may not be much different than us after all. Researchers let 25 AI avatars — otherwise referenced as “generative agents,” software agents that can simulate “believable human behavior” — with unique and personalized identities move into a virtual town.
Inspired by the video game “The Sims,” these avatars lived in a small town community called Smallville. The generative agents that were implemented allowed the avatars to conduct behaviors such as waking up, cooking breakfast and commuting to work. Artists were able to paint and writers wrote, according to the study. They formed opinions, reflected on their day and planned ahead — some were even failing to remember certain events that happened during the day.
The architecture utilizes a “turbo version” of gpt3.5, the model behind ChatGPT, to get the agents to perform certain tasks.
Here’s One Way Artificial Intelligence Could Earn the Trust of Americans
A decade from now, the much-maligned tech just might save your lifeOne avatar, Isabella Rodriguez, threw a Valentine’s Day party at a cafe in the town. Avatars sent out invitations to their friends and asked each other on dates to the function. They also coordinated what time to show up and followed through on their arrival time. Another avatar called Maria also helped Isabella decorate the cafe before the party began.
Isabella and Tom Moreno, another avatar in the Smallville community, held a conversation about the upcoming election. Their conversation even included some small town gossip. “I’m still weighing my options, but I’ve been discussing the election with Sam Moore,” Isabella said. “What are your thoughts on him?”
“To be honest, I don’t like Sam Moore,” Tom said. “I think he’s out of touch with the community and doesn’t have our best interests at heart.”
The study concluded by saying that this provided avatars with the ability to “deepen its understanding of itself and environment through reflection.”
In a report from Business Insider, Michael Wooldridge, a professor who studies AI at Oxford University, said this study reveals that these are “baby steps” toward “artificial general intelligence,” which is the ability for AI to “display complex human behaviors like consciousness.” And while there’s still more research to be done, the researchers also said that going forward, agents may just be capable of creating believable behavior and taking on different roles in interactive applications.
Thanks for reading InsideHook. Sign up for our daily newsletter and be in the know.