Simulation: the new Scaling Law — Joon Sung Park, Simile AI

Simile AI is a company focused on simulative artificial intelligence, having raised a two billion dollar Series B funding round backed by prominent industry figures and investors. The venture builds upon foundational research like the 2023 Smallville paper on generative agents, which demonstrated that AI characters could remember, plan, socialize, and exhibit emergent behaviors. Simile currently runs tens of millions of simulations for Fortune 100 clients, achieving 85 to 99 percent accuracy compared to traditional human focus groups.

The core of this technology involves modeling human behavior to create digital twins and foundation models of people. Standard frontier language models often fail to capture real human behavior because they rely heavily on web data, which reflects what people say online rather than what they actually do in daily life. To bridge this gap, Simile incorporates three types of data: long-form qualitative interviews about life stories and personal contexts, observational and transaction data for baseline statistics, and randomized controlled trials that capture the causal mechanisms behind decision-making.

This approach enables organizations to test products and policies before deploying them in the real world. Rather than simply predicting future outcomes, simulations allow decision-makers to understand how to shape the future, test counterintuitive paths, and study emergent behavior across entire societies. The long-term ambition for this technology includes simulating all eight billion people on Earth to help address complex global challenges like climate change, economic policies, and democratic instability.