Mirror Particle wants to model human behavior. Care?

TechCrunch reports that Mirror Particle, a two-year-old San Francisco startup, is building what it calls a "world model" of human behavior. The goal is to predict why people do what they do, and what they will do next, better than a chatbot can.

The company's CEO, Abhivyakti Ahuja, thinks the usual approach is broken. Most prediction tools rely on large language models that are prompted or fine-tuned to act like a target demographic. Ahuja told TechCrunch that approach is "like bringing a super soaker to Niagara Falls." LLMs model written language, she says, but humans are made of visual perception, spatial reasoning and social intelligence. Relying on them gives you insights based on what humans don't notice, which is beside the point when the job is predicting human behavior.

Mirror Particle's alternative is a foundation model built from scratch. Ahuja describes it as simulating why humans do what they do and how behavior changes over time. The company combines client customer data, current events, pop culture, social media and more, then tracks what people actually do rather than what they tell a survey. It also wants to capture "the changing person" rather than a static snapshot. If someone isn't changing, she said, that's also a signal.

TechCrunch notes that Mirror Particle is not alone. Simile raised $200 million at a $2 billion valuation. Aaru raised $88 million at a $1 billion valuation. Humans& announced a $480 million seed round in January at a $4.48 billion valuation and launched a product called Persimmon. Money is flowing into the idea that predicting humans is the next big thing.

What this means for you

Not much yet. Mirror Particle sells to brands, not consumers. In one early pilot, a pet food brand learned that the imagery on its packaging did not matter; the problem was that the brand felt mass market and cheap. That insight helps a marketing team. It won't change your phone settings.

The company says it is close to closing its first venture round and will be at TechCrunch Disrupt in San Francisco on Oct. 13-15 as part of the Startup Battlefield 200. If you are not at Disrupt, you can ignore this until there is a consumer product. When there is, we will tell you whether it's Buy, Wait or Skip.

• Abhivyakti Ahuja
• AI
• human behavior
• Mirror Particle
• TechCrunch Disrupt