Israeli startup 42AI has come out of stealth with a bold sci-fi claim and $62.5 million in seed funding: it’s building a “behavioral world model” that can predict what a person will do before they decide — even in completely new situations.
42AI logo (Credits: TechCrunch)
The company is initially running its predictive system at a handful of financial brokerages, where it says the model can forecast how individual traders will respond to volatile market conditions.
How 42AI’s Behavioral World Model Works
Rather than extrapolating from a person’s past actions, 42AI says its model tries to uncover internal motivations. It examines how people react to seemingly unrelated events — for a trader, that might include KYC onboarding, support calls, money transfers, and voice sentiment — and converts those observations into mathematical representations of psychological traits.
Those traits feed a transformer-based world model that maps out the probabilities of different decisions under different circumstances, with a separate language model generating explanations for each prediction. CEO and co-founder Amit Gal-Or argues people have underlying psychological tendencies that persist even as circumstances change, and those tendencies are what the system measures.
“A week from now, you might be a different person, but your tendencies will always play out in certain areas to a probability or distribution,” added co-founder and CTO Ehud Ezra.
Three Years Of Development With Psychologists And A Nobel Laureate
42AI spent three years developing the approach, combining psychological research with neuroscience and genetics. The team studied thousands of people through psych assessments and behavioral scenarios before adding government and biobank datasets, and consulted psychology and PTSD experts alongside Dan Shechtman, the 2011 Nobel Prize winner in Chemistry.
The model is trained so a person’s underlying tendencies hold steady in simulations — if a usually risk-averse person suddenly acts like a risk-taker in the model, “the target is off,” Ezra said, describing how training rewards are structured.
Whether a behavioral world model can truly generalize to strangers in novel situations is the big open question — but $62.5 million says investors are willing to find out.
Also Read: Omen AI Acquires Aquatic Labs In $16 Million Deal To Fix The Data Center Water Problem | Ernd Launches The Credit Card For The AI Stack: 4% Back On AI Vendor Spend, No Annual Fee
Source: TechCrunch — read the full story there.
