Doom or Bloom
问题 1
你认为AI对我们的未来意味着什么？为什么？
I expect AI to matter enormously eventually, but not to transform the whole economy overnight. AI is a general-purpose technology: realizing its value requires complementary capital, redesigned organizations, new workflows, legal adaptation, and time. Impressive personal usefulness or rapid model progress does not imply equally rapid economy-wide reorganization. My default is therefore large effects unfolding over decades, rather than a sudden concentrated takeoff.
The central comparison is not “dangerous AI versus perfect control.” It is AI risks versus the costs and failure modes of the institutions proposed to control AI. Calling an indefinite regulatory regime a “pause” does not show that alignment will soon be solved or that regulators will use their power well. Under present governance, I prefer ordinary law and liability to politically driven AI-specific restrictions, while still supporting investigation of concrete safety problems.
I also think discussion is oddly asymmetric about values. Human cultures and descendants can drift too; that risk should be compared with AI value drift rather than treated as a fixed human standard confronting alien machines. Current language models look unusually prosocial to me. Conditionally, human-level AI or emulations competing and adapting could even help preserve functional cultural variation. But alignment rules and AI-rights regimes could themselves freeze particular values and suppress experimentation. So AI’s future depends not just on machine capability, but on institutional competition, adaptation, and whether our attempted safeguards become larger hazards than the problems they target.
