Simile AI Raises $2B: Simulation as the New Scaling Law
Latent Space · rss · 2026-08-22
The Latent Space podcast interviews Joon Sung Park, co-founder of Simile AI, exploring the rise of "simulation" as a new AI paradigm. Simile AI recently closed a $2B Series B (led by GreenOaks and Index Ventures, backed by Fei-Fei Li and Andrej Karpathy), running tens of millions of simulations for Fortune 100 clients with 85–99% accuracy.
Key Insights & Technical Approach:
- From Smallville to Behavioral Foundation Models: Joon led the famous "Smallville" generative agents paper. Simile is now building foundation models of human behavior, creating "digital twins" that replicate real human behaviors and attitudes with 85% accuracy.
- Why Simulation?: Frontier models trained on web text capture "what people say" rather than "what people actually do." Simile uses long-form interviews, observational data, transaction data, and Randomized Controlled Trials (RCTs) to train models on the causal mechanisms behind decisions.
- "Social Physics": To simulate humans accurately, models must replicate human biases and irrationalities, not just rationality. This requires more than prompting; it may involve adjusting model weights.
- Applications: Testing products and policies before deployment (e.g., replacing expensive human focus groups), finding counterintuitive paths to outcomes, modeling emergent behavior, and addressing macro issues like climate change or democratic instability.
- Scale & Ambition: Discussion covers the possibility of simulating all 8 billion people and the compute economics of simulating the world (potentially requiring entire data centers).
- Relation to AGI: Simulation and AGI may be the twin technologies of advanced civilizations.
More from AGI Musings
- Mech Interp Veteran Rethinks Alignment, Misgeneralization, CoT Oversight — akbirthko · 2026-10-16
- GitCoin founder on Nick Land's technocapital: capitalism as a self-assembling machine intelligence — khademinori · 2026-10-16
- AI-driven job losses could turn $10T consumer debt into a $1T banking hole — churchkey · 2026-10-16
- Into the Unknown: an AI agent benchmark of 15 open research problems from US national lab scientists — Jeande_d · 2026-10-16
- Soon your AI will meet their AI — and humans without agents can't compete — danfaggella · 2026-10-16
- Analyst: in a year, models 2-3x better across the board at 10x lower cost, token use up 10-20x — DeryaTR_ · 2026-10-16