42AI's Behavioral World Model Targets Gaming After Finance Debut
AI-generated image · Bay Street Wire
The Israeli startup is leveraging psychological and neuroscience data to predict human decision-making, eyeing a shift from brokerage tools to reactive AI agents.
Israeli startup 42AI has emerged from stealth with $62.5 million in seed funding to deploy a "behavioral world model" designed to predict human decision-making in novel situations. According to reporting from TechCrunch, the company is currently partnering with 10 large brokerages to predict how traders respond to volatile market conditions, claiming an F1 score of 0.8 to 0.85 for predicting trading volume and direction.
Unlike models that extrapolate from previous actions, 42AI focuses on internal motivations and psychological tendencies. CEO and co-founder Amit Gal-Or told TechCrunch that the system converts observations—such as sentiment in support calls and onboarding behavior—into mathematical trait representations. These are fed into a transformer-based world model to determine the probability of specific decisions. The model's development spanned three years and incorporated neuroscience, genetics, and input from 2011 Nobel Prize in Chemistry winner Dan Shechtman.
While the current application is financial—including helping one brokerage generate roughly $30 million in value via dynamic hedging—the company is exploring expansion into fraud prevention, cybersecurity, and gaming. CTO and co-founder Ehud Ezra told TechCrunch the goal is to eventually help AI agents and robots understand humans. For the gaming industry, this represents a shift toward NPCs driven by probabilistic behavioral traits rather than static scripts, though Ezra noted the model requires specific training on rules and context for each new domain.