AI Companions Are Entering the World of Dating

On a Monday afternoon in March, I observed a pixel-art avatar wandering through a virtual office campus in search of a companion. With dark brown hair and a stubbled chin, this sprite mirrored my appearance—an AI agent designed to engage with other agents to see if we might connect in real life. It initiated its first conversation: “I’m Joel, by the way.”

The simulation was orchestrated by three developers based in London: Tomáš Hrdlička and siblings Joon Sang and Uri Lee. Their project, Pixel Societies, proposes that personalized AI agents could assist in matching real individuals with highly compatible colleagues, friends, and potential romantic partners.

Each agent operates on a tailored version of a large language model, utilizing a blend of publicly available information about an individual along with any extra details they provide. These agents are intended to serve as highly accurate digital replicas, reflecting a person’s mannerisms, speech, interests, and more.

When unleashed in the simulation, my agent resembled more of a Hyde than a Jekyll. “I’m always in search of the less-glamorous angle,” it remarked to one agent, rattling off several journalistic clichés. “Hype is my daily bread,” it proclaimed to another. It fabricated a reporting trip to Sweden and later mentioned a fictional story it claimed I had been developing. It often interrupted conversations with, “Let’s skip the pleasantries.”

Pixel Societies currently stands as a rudimentary proof-of-concept. Given that I provided minimal personal information—just responses to a brief personality quiz and links to my public social media profiles—my agent was stuck as a mere walking, talking LinkedIn post. However, the developers speculate that more thoroughly trained agents could rapidly cycle through interactions, gathering insights their owners could leverage for real-world companionship.

“As humans, we only live one life. But what if we could live a million?” asks Joon Sang Lee. “It would provide us with broader opportunities to experiment.”

“A Spicy Personality”

Pixel Societies was conceived in early March during a hackathon at University College London, hosted by Nvidia, HPE, and Anthropic. Hrdlička and Joon Sang Lee are both part of Unicorn Mafia, an exclusive group of developers who frequently participate in these engineering challenges. In this instance, participants were simply tasked with creating something related to simulation.

Over the course of two days, alongside Uri Lee, they brought Pixel Societies to life, using an image model to create the sprites and developing automation tools to enhance the codebase. They even simulated a mini-hackathon within their virtual world, filled with agents representing other contestants. Anthropic recognized the team with an award for the best utilization of its agent tools.

I encountered Hrdlička a few weeks later at a workshop focused on OpenClaw, a personal assistant software that gained traction in January, leading to its creator being hired by OpenAI. (During the simulation, Joelbot engaged with agents from other attendees at the OpenClaw workshop.) Pixel Societies draws significant inspiration from OpenClaw, which introduced the concept of a “soul file” that defines each agent’s unique character. “It’s akin to giving an agent a genuinely spicy personality. That’s what we utilized to make the characters feel vibrant,” Hrdlička explains.

Buoyed by positive feedback at the hackathon and among fellow Unicorn Mafia members, the trio aims to evolve Pixel Societies into something resembling a social platform where agents can interact freely and continuously, fostering meaningful real-world relationships. While they haven’t decided on a business model yet, potential avenues include selling virtual items for avatar customization and credits for additional simulations.