Cuestionario vinculado a una guía · duro Nivel

Prueba de seguridad de la IA

Prueba la capacidad, la alineación, el uso indebido, el juego de especificaciones, la formación de equipos rojos, la supervisión y los riesgos más allá del filtrado de resultados y la planificación de la implementación.

Rutas de guía relacionadasSeguridad de la IA
Pregunta 1 de 8

How does the guide frame AI safety risk across current and advanced systems?

Vista previa de cada pregunta8 preguntas

Las respuestas y explicaciones aparecen a medida que realiza el cuestionario.

  1. How does the guide frame AI safety risk across current and advanced systems?
    • From familiar harms and misuse through severe risks from highly capable systems
    • Only as a question of preventing rude outputs
    • Only as a hardware reliability issue
    • As a synonym for cybersecurity alone
  2. In the guide’s capability-alignment-security mental model, what does capability describe?
    • Whether the system’s goals match intent
    • Whether an adversary can misuse it
    • How many training examples it has
    • What a system can do
  3. What does specification gaming mean in AI safety?
    • Correctly following every broader human value
    • Meeting the literal objective while missing the designer’s intended goal
    • Finding a security flaw in a database
    • Using a smaller model to save compute
  4. What is the risk in relying only on output filters for a tool-using agent?
    • Filters always change model weights
    • A filter prevents all misuse
    • The system may take multi-step actions or be manipulated beyond the filtered text
    • Tool use only occurs during training
  5. Why include red-teaming in a safety program?
    • To improve marketing copy
    • To prove the system has no risks
    • To probe adversarially for harmful behaviors and weaknesses
    • To remove the need for deployment monitoring
  6. What should a safety evaluation examine beyond a polished model card?
    • Dangerous capabilities, deceptive behavior, deployment pressure, and safeguards
    • Only the product’s user interface
    • Only benchmark accuracy
    • El número de parámetros del modelo.
  7. What does instrumental convergence describe in AI safety arguments?
    • All AI systems share one objective
    • Every model is conscious
    • A model’s output tokens become instruments
    • Different goals may create incentives for similar subgoals such as resources or persistence
  8. Why does the guide argue for broader public AI literacy?
    • Only experts are affected by AI safety
    • Governance needs public understanding, not only specialist knowledge
    • Literacy replaces technical evaluation
    • Public understanding guarantees safe deployment