AI capability control

<p>In the field of <a href="/facts/Artificial_intelligence/lJGauQwX">artificial intelligence</a> (AI) design, AI capability control proposals, also referred to as AI confinement, aim to increase our ability to monitor and control the behavior of AI systems, including proposed <a href="/facts/Artificial_general_intelligence/vbxwaq8t">artificial general intelligences</a> (AGIs), in order to reduce the danger they might pose if <a href="/facts/AI_alignment/Blgof0uy">misaligned</a>. However, capability control becomes less effective as agents become more intelligent and their ability to exploit flaws in human control systems increases, potentially resulting in an <a href="/facts/Existential_risk_from_artificial_general_intelligence/j7oxWDYY">existential risk from AGI</a>. Therefore, the Oxford philosopher <a href="/facts/Nick_Bostrom/kyNalABj">Nick Bostrom</a> and others recommend capability control methods only as a supplement to <a href="/facts/AI_alignment/Blgof0uy">alignment methods</a>.
</p>

AI capability control open-in-new

AI capability control