Jev HighwayEnv: 60 Seconds Without a Crash
“I'll send Jev the cars' positions and speeds and have it choose whether to change lanes or slow down. How hard can that be?” That was the idea. Instead, the car crashed almost immediately. When I added braking, it stopped and stayed there. When I added acceleration, it crashed again. This is the story of getting it to drive for 60 seconds—roughly 1.20 km—without a collision, on a three-lane…
Jev is an AI model specifically designed for decision-making, not for generating free-form text like a language model. Developed by TypeSafe AI, it functions as a System One model, which takes a given state, evaluates it, and returns typed answers along with associated probabilities that can be directly utilized by software. The name "Jev" reflects the model's design, inspired by Daniel Kahneman's concept of System 1 - a fast, intuitive mode of thinking.
Unlike other models, Jev is trained using Reinforcement Learning for Calibrated Decisions (RLCD), allowing developers to use its outputs to determine the subsequent actions of an agent. The model receives a state and a set of questions, and returns evaluations and probabilities for each possible course of action. For instance, a developer could ask Jev whether a support request is urgent, and receive a high probability of it being urgent.
Jev's design enables it to quickly respond to complex decision-making scenarios, making it suitable for applications requiring real-time decision-making.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
