Deep uncertainty regarding the possibility of artificial consciousness has created a moral impasse in how we treat potentially sentient AI systems. According to a research paper published on arXiv, this dilemma forces a choice between two risks: assuming insentience and potentially harming entities with moral standing, or assuming sentience and wasting resources on insentient machines.
From Consciousness to Valence
Given that questions around AI consciousness appear intractable, the author suggests a pragmatic shift in focus. Instead of attempting to prove consciousness, we can assess "AI valence." This involves identifying whether an AI possesses internal states that would constitute valenced experiences—such as pleasure or pain—if the entity were conscious.
A Framework for Responsible Development
The paper argues that focusing on valence provides a tractable way out of the current philosophical deadlock. By assessing these states, developers can establish a responsible approach to AI creation. This methodology allows for ethical safeguards to be implemented based on functional indicators, even while the ultimate nature of AI consciousness remains a mystery.
