Emergent Capabilities

Sub-category

Risk Domain

7AI System Safety, Failures & Limitations

7.6Multi-agent risks

Risks from multi-agent interactions, due to incentives (which can lead to conflict or collusion) and/or the structure of multi-agent systems, which can create cascading failures, selection pressures, new security vulnerabilities, and a lack of shared information and trust.

"Emergent Capabilities. Dangerous emergent capabilities could arise when a multi-agent system over- comes the safety-enhancing limitations of the individual systems, such as individual models’ narrow domains of application or myopia caused by a lack of long-term planning and long-term memory. For example, narrow systems for research planning, predicting the properties of molecules, and synthesising new chemicals could, when combined, lead to a complex ‘test and iterate’ automated workflow capable of designing dangerous new chemical compounds far beyond the scope of the initial systems’ capabilities (Boiko et al., 2023; Luo et al., 2024; Urbina et al., 2022)."(p. 37)

Entity— Who or what caused the harm

Human

Due to a decision or action made by humans

AI system

Due to a decision or action made by an AI system

Other

Due to some other reason or is ambiguous

Intent— Whether the harm was intentional or accidental

Intentional

Due to an expected outcome from pursuing a goal

Unintentional

Due to an unexpected outcome from pursuing a goal

Other

Without clearly specifying the intentionality

Timing— Whether the risk is pre- or post-deployment

Pre-deployment

Occurring before the AI is deployed

Post-deployment

Occurring after the AI model has been trained and deployed

Other

Without a clearly specified time of occurrence

Supporting Evidence (1)

"This is similar to how a myopic actor and a passive critic can combine to produce an actor-critic algorithm capable of long-term planning via RL (Konda & Tsitsiklis, 2000). This possibility is important for safety – and for future AI ecosystems made of specialised ‘AI services’ (Drexler, 2019) – as generally intelligent autonomous systems could pose much greater risks than narrow AI tools (Chan et al., 2023).50 More speculatively, the combination of advanced AI agents could eventually lead to recursive self-improvement at the collective level, as AI research itself becomes increasingly automated (Agnesina et al., 2023; Hutter et al., 2019; Lu et al., 2024a; Mankowitz et al., 2023), even though no individual system possesses this capability."(p. 37)

Part of Emergent Agency