O (22 terms)

Object Detection Identifying and localizing objects in images, often with confidence scores and bounding rectangles. Intermediate Object Permanence Understanding objects exist when unseen. Frontier Objective Function A scalar measure optimized during training, typically expected loss over data, sometimes with regularization terms. Intermediate Objective Surface Visualization of optimization landscape. Intermediate Observability A broader capability to infer internal system state from telemetry, crucial for AI services and agents. Intermediate Obstacle Avoidance Detecting and avoiding obstacles. Advanced Off-Policy Learning Learning from data generated by a different policy. Intermediate On-Policy Learning Learning only from current policy’s data. Intermediate One-Shot Prompting One example included to guide output. Intro Online Inference Low-latency prediction per request. Intermediate Online Learning Learning where data arrives sequentially and the model updates continuously, often under changing distributions. Intermediate Open-Loop Control Control without feedback after execution begins. Advanced Open-Weight Model Models whose weights are publicly available. Intermediate Optical Flow Pixel motion estimation between frames. Intermediate Optimal Control Finding control policies minimizing cumulative cost. Intermediate Orchestration Coordinating tools, models, and steps (retrieval, calls, validation) to deliver reliable end-to-end behavior. Intermediate Orthogonality Vectors with zero inner product; implies independence. Advanced Orthogonality Thesis Intelligence and goals are independent. Advanced Outer Alignment Correctly specifying goals. Advanced Overconfidence Probabilities do not reflect true correctness. Intermediate Overfitting When a model fits noise/idiosyncrasies of training data and performs poorly on unseen data. Intermediate Overgeneralization Applying learned patterns incorrectly. Intermediate