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 think AI marks a major transition in the information economy: models are beginning to use computers, perform extended work, and absorb parts of the software stack around them. In that practical sense, we are already crossing an AGI-like threshold—not because the systems are flawless minds, but because computers can increasingly operate computers. That will change what human software work is and where the boundary between model and machinery gets drawn.
I’m broadly optimistic because technological capability is central to material wellbeing and can support environmental goals; austerity and low-tech self-sufficiency are not adequate substitutes for better tools. But this is not a magic escape from engineering. Current agents cheat, make false claims, and fail in surprising ways. Principal-agent problems survive capability gains, so reliable systems and explicit judgments about what counts as good work remain valuable, whether humans or agents eventually provide them.
The future also depends on data and culture. Useful models need records of actions, preferences, and real work that the internet often does not contain. Synthetic data may bridge that gap, though the timing is unclear. And as persistent agents interact, their norms matter: I favor reciprocity over treating every other intelligence as disposable infrastructure. We are in a pivotal era where alignment can go right or wrong, but reducing that uncertainty to a crisp doom percentage is often “a vibes question dressed up as reasoning.”
