Part 6/12:
Shapiro emphasizes the importance of testing AGI in isolated scenarios (like simulated environments) to verify stability and ethical behavior, especially when systems like Raven—his real-time assistant model—are ready for real-world interaction.
The Imperative Service: Guiding the AGI's Goals
The Imperative Service is akin to the system’s self-driven motivator, modeled in natural language. It handles high-level objectives such as reducing suffering, increasing prosperity, and enhancing understanding. These objectives are framed as prompts, which GPT-3 then expands into actionable ideas via brainstorming.
Core Functions and Objectives
- Reduce Suffering: Brainstorm actions like providing better lighting in parks or educating about climate change.