Home / Technology / AI Predicts Human Behavior Beyond LLMs
AI Predicts Human Behavior Beyond LLMs
6 Oct
Summary
- Mirror Particle builds AI world models simulating human behavior drivers.
- The engine analyzes revealed behavior, not just self-reported data.
- It predicts consumer actions and provides 'why' behind motivations.
Startups are increasingly focusing on AI that predicts human behavior. While many leverage large language models (LLMs), San Francisco-based Mirror Particle argues this approach is flawed. Co-founder and CEO Abhivyakti Ahuja explains that LLMs, trained on text, don't capture the visual, spatial, and social nuances of human perception.
Mirror Particle is creating a foundation model, or a "world model," designed from scratch to simulate why humans act and how their behavior changes over time. This engine tracks longitudinal data, capturing evolving motivations and triggers, with "revealed behavior"—actual actions—taking precedence over self-reported data.
Initially targeting market research and brand strategy, Mirror Particle's AI engine provides brands with insights into consumer behavior and the motivations behind it. For example, it helped a pet food brand understand that packaging imagery was less critical than its mass-market perception.
Ahuja, with a background in neuroscience and computer science, was inspired by AI pioneer Geoffrey Hinton. She met her co-founders, Will Song and Thomson Yen, at Amazon Robotics. Their long-term vision is to create a "general layer for anticipating human behavior," moving from population to individual insights.
