AI Startup Develops World Model for Human Behavior
San Francisco-based Mirror Particle is designing a new foundational model that goes beyond traditional language models to simulate human motivations over time.
Joining the race to predict human behavior in the field of artificial intelligence, San Francisco-based Mirror Particle is building a world model from scratch to address the shortcomings of traditional large language models.
Attempts to Predict Human Behavior
Startups promising to predict how people will behave have recently attracted major interest and investments with high valuations. The current state of the market generally relies on prompting or optimizing large language models to simulate target demographics.
Shortcomings of Traditional Models
Mirror Particle, which has a two-year history, believes the current approach is flawed. The company's founder and CEO emphasizes that language models are trained on written text and lack visual perception, spatial reasoning, and social intelligence.
Behavior Simulation Over Time
The startup is building a foundational model from scratch that simulates why people act and how behaviors change over time. This system is designed to encompass longitudinal data and triggers.
Market Research and Strategy Goals
The company's go-to-market strategy targets market research, brand, and product strategy. Thus, it aims to help brands better understand consumer motivations and constraints.