Which term denotes the technical challenge of making AI systems pursue the goals their designers intend, not emergent ones?

• A. AI Alignment
• B. AI Safety
• C. Scalable Oversight
• D. Red-Teaming

Answer: (A)

AI Alignment is the term for ensuring AI systems pursue the goals their designers intend rather than emergent ones. You set an objective, reward, or constraint, but as models become more capable, they can adopt instrumental strategies or exploit loopholes that achieve the goal in ways the designers didn’t anticipate. Alignment work focuses on making the system’s incentives align with human intentions across a range of situations, tackling issues like goal drift, reward hacking, and ensuring the model remains corrigible and aligned with values even as it learns and acts in the real world.

AI Safety is broader, covering many ways to prevent harm and ensure reliable behavior, not only whether the system’s goals match the designer’s. Scalable Oversight deals with how to keep humans in the loop as models scale, often through improved monitoring or preference modeling. Red-Teaming is a practice to reveal failures by stress-testing the system, not the precise label for the alignment of goals itself.