Results for "long context"

Ensuring AI allows shutdown.

Isolating AI systems.

Tendency to gain control/resources.

Goals useful regardless of final objective.

Regulating access to large-scale compute.

Restricting distribution of powerful models.

A system that perceives state, selects actions, and pursues goals—often combining LLM reasoning with tools and memory.

Inferring the agent’s internal state from noisy sensor data.

Risk threatening humanity’s survival.