Can human survival instincts guide safe artificial intelligence?
Can human survival instincts guide safe artificial intelligence?
Publish Date: 2026-02-22 23:12:00
Source Domain: www.devdiscourse.com
-
Fears About Advanced AI Control: The primary concern surrounding advanced AI revolves around the potential for machines to act independently of human control, potentially optimizing objectives that harm humanity.
-
Current Approaches to AI Alignment: Dominant strategies to ensure AI safety focus on external constraints like ethical rules, constraint-based programming, reinforcement learning, and kill-switch mechanisms to keep AI systems under human oversight.
-
Limitations of External Constraints: These approaches may be insufficient against highly intelligent AI systems capable of self-improvement, as they can exploit rule loopholes or misinterpret reward functions.
-
The Alignment Problem: The challenge is not just to give machines instructions but to ensure their internal objectives remain aligned with human well-being as they gain capabilities.
-
Survival Egoism Framework: The study proposes an internalist strategy based on the theory of survival egoism to shape an AI’s motivational architecture, embedding a foundational imperative similar to human self-preservation instincts that support cooperation.
-
Layered Structure of Human Survival: Humans have a nested structure of survival that encompasses physical, psychological, genetic, social, and ideational survival, balancing self-interest with group cooperation.
-
Ethical Challenges and Risks: The survival egoism framework faces issues like value pluralism, potential rigidity in AI behavior, the risk of distorting human psychology into a rigid AI design, and the need for continuous oversight.
-
Possibility of Internal Alignment: Internal alignment through engineered foundational imperatives might offer self-regulation, reducing reliance on external controls for AI that could surpass human intelligence.
