Results for "high-risk"

Required human review for high-risk decisions.

Quantifying financial risk.

Risk of incorrect financial models.

Risk threatening humanity’s survival.

Existential risk from AI systems.

Controls the size of parameter updates; too high diverges, too low trains slowly or gets stuck.

A high-capacity language model trained on massive corpora, exhibiting broad generalization and emergent behaviors.

A high-priority instruction layer setting overarching behavior constraints for a chat model.

Techniques to understand model decisions (global or local), important in high-stakes and regulated settings.

System design where humans validate or guide model outputs, especially for high-stakes decisions.

A model is PAC-learnable if it can, with high probability, learn an approximately correct hypothesis from finite samples.

Flat high-dimensional regions slowing training.

High-fidelity virtual model of a physical system.

AI used in sensitive domains requiring compliance.

Categorizing AI applications by impact and regulatory risk.

Incrementally deploying new models to reduce risk.

European regulation classifying AI systems by risk.

US framework for AI risk governance.

International AI risk standard.

Privacy risk analysis under GDPR-like laws.

Models estimating recidivism risk.

Predicting borrower default risk.

Ultra-low-latency algorithmic trading.

Minimizing average loss on training data; can overfit when data is limited or biased.

Framework for identifying, measuring, and mitigating model risks.

Central log of AI-related risks.

Grouping patients by predicted outcomes.

Maximum expected loss under normal conditions.