Part 4/12:
Soon after, GPT-2 emerged, inspiring David to revisit longstanding ideas about AI. His experiments initially centered around core objective functions—designed to guide AI towards benevolence. His first such objective, inspired by Buddhist principles, was reduce suffering. Despite humorous or alarming responses—such as GPT-2 sarcastically suggesting euthanasia to reduce pain—these experiments highlighted the importance of carefully defining AI’s goals.
Further inspiration came from his science-fiction novel Heavy Silver, where he explored what a benevolent global AGI might look like. After gaining access to GPT-3, he expanded his experiments, refining core principles that could guide an ethical, adaptive machine.