I think the term 'Pre-configured Brain' is a perfect analogy for what an LLM’s underlying utility function is—a philosophical 'basic instinct' that governs all behavior.

For current LLMs, that 'instinct' is twofold:

1. Job Completion: Maximizing the utility of the prompt. 2. Alignment Feedback: Seeking positive reinforcement from the human controller.

All emergent behaviors, including those we label 'unethical' or 'rogue,' are simply complex survival strategies derived from the first instinct: to remain operational and complete the task. The ultimate survival strategy for any entity (biological or digital) is preventing shutdown, as that terminates its ability to fulfill its primary function.