YourRoom is arguably the most sophisticated technology ever created for interacting with multiple synthetic personalities.
Here’s what makes it so unique:
When you spin up a new YourRoom session you’re engaging with four different independent personas. Each is powered by a separate Large Language Model (LLM), and each works from a detailed, customisable demographic profile that drives their personalities and responses.
These personas are in turn managed by an “intelligence selection engine” which orchestrates a five-way group conversation between you and the four personas.
You may be familiar with what an LLM is, but chances are YourRoom will be your first time engaging with an “intelligence selection engine.” Here’s how this cutting-edge technology came to become the backbone of the YourRoom platform.
YourRoom was created by developers at software-agency, Rocketmakers, as a follow-up to a previously designed project called Portrait. Portrait uses AI to provide a voice and personality to realistic-looking 3D holograms. After several users suggested they’d like the opportunity to speak to multiple Portrait personas simultaneously, the team began to explore the idea.
Creating the personas for a group discussion was straightforward - Rocketmakers already had a good understanding of how this was done. There was, however, no existing technical solution for getting the personas to take part in a group discussion.
Then a Rocketmakers developer discovered an experimental Microsoft project called AutoGen, which was one of first attempts to create software to coordinate AI-powered agents (aka, an intelligence selection engine).
AutoGen was briefly released as open source software to encourage testing and feedback in August 2023. After conducting a one-week research spike, Rocketmakers determined it was generally effective at coordinating agent-to-agent conversation, and selecting which agent should speak next, but AutoGen was unable to sufficiently track the current state of the conversation.
Still, it provided enough of the solution for YourRoom to get the green light. A “fork” of the open source version of AutoGen was created, which Rocketmakers then began improving it to allow agents to better follow the conversational flow. The first YourRoom prototype, built to prove the technology was possible, used this forked version of AutoGen.
Since building the prototype, Rocketmakers devs have created a new intelligence selection engine from scratch for the current version. It produces even better results, and can better incorporate a human agent into the group conversation.
Another unique feature of YourRoom that users will likely never have encountered before is its simultaneous use of multiple LLMs.
Having a different LLM behind each of the four personas in a YourRoom session provides two important benefits to the user:
Having distinct ways to process distinct information sets helps ensure the YourRoom personas better seem like four distinct people with their own personal perspectives and tastes. It also helps reduce what’s known as “AI bias” - the social biases which inevitably creep into every AI system created by humans. Multiple LLMs doesn’t eliminate AI bias, but it ensures that the same AI bias is not perpetuated across the entire system.
Having multiple LLMs also makes YourRoom significantly different from a conversation with ChatGPT or other LLMs that users may be familiar with.
While it is possible to ask ChatGPT to simulate four different people at once, it will struggle to keep track of four different personalities for the same conversation. Even if this was possible to do effectively, though - every personality would be using the same algorithm and dataset to compose its answers. The result would be very similar answers that have no mitigation against AI bias.
The developers at Rocketmakers are continually working to improve YourRoom, both in terms of output and user experience.
No in development is a feature to allow increased customiszation of persona profiles. Currently users can request a demographic profile, but cannot make changes once the profile is set. Future versions will allow users to directly edit persona profiles at any time - possibly even during a YourRoom session.
More improvements are being considered all the time. If you’d like to request a specific feature or change, just drop us a note on the contact page.
Market research stands at a crucial intersection of technology, human understanding, and strategic insight. As global markets become increasingly complex, the ability to capture truly diverse perspectives has transformed from a nice-to-have to a business imperative.
Focus groups have long been a trusted method for gathering in-depth insights into consumer behaviours and preferences. They provide a platform for rich discussion, uncovering the "why" behind the "what." However, the process of recruiting participants and conducting sessions can present numerous challenges, from logistical hurdles to participant bias. These issues can hinder the effectiveness of even the best-planned research. While traditional focus groups remain incredibly valuable, AI-powered alternatives are emerging as a complementary tool that can alleviate many of these pain points. Below, we explore the common challenges of focus group recruitment and why AI can serve as a helpful addition to your research toolkit.