THE GOD TEST

Are we headed for AI heaven or AI hell?

Wright (The Evolution of God, 2009) examines ideas from Jesuit philosopher Pierre Teilhard de Chardin to Dario Amodei, founder of Anthropic, weighing arguments for and against rapid AI development, as the capabilities of large language models (LLM) accelerate to superhuman-like levels. When AI becomes a global brain, is this essentially a godlike entity in control? After explaining the workings of LLMs, he dives into AI development, a reflection of the range of recorded human thinking. Summarizing books on the subject—a helpful reading list for the AI-curious—he explores the conflicting views of pioneering AI “accelerationists” vs. “doomers,” including Geoffrey Hinton, Peter Thiel, and Nick Bostrom, who envisions a world government controlled by AIs. The ideas of Teilhard, particularly the “noosphere”—a “brain of brains” or world consciousness created by the actions of all human brains—loom large in Wright’s discussion, connected to current LLMs. He observes, “If a large language model is in some sense a crystallization of the noosphere, why should you expect it to be more evolved, more enlightened, than the noosphere itself—which is to say, than us?” Instead, “technological evolution strongly encourages us to—for our own sakes—become better people.” Wright’s survey of AI experts is somewhat meandering but presents many consequential debates on AI systems as shapers and mirrors of human values. His meditations convince himself, and perhaps readers, that “The more God-like AI is, the more it will put the fear of God in us, and the better we’ll get at cooperating to tame its ferocity and reap its benefits.”

A compelling case for bending AI closer to human values by changing ourselves.