Before solving alignment on AI we must solve alignment of the oligarchs.
讨论提出了一个问题：在解决寡头们的"对齐"问题之前，人工智能的"对齐"是否能真正实现。文中引用了一个寓言故事：一个人宁愿自损一眼，也不愿邻居获得双倍的好处，以此来比喻一种普遍存在的担忧，即有权势者所定义的"对齐"可能仅仅意味着服从他们个人。核心的担忧在于，当自私和不希望他人成功的心理普遍存在时，如何才能就一个真正惠及所有人的"对齐"定义达成共识。
时间与来源
时间显示为 UTC
显示时区：UTC
本地时区尚不可用，暂时显示 UTC。
发布当时偏移：UTC+02026年9月11日 22:03 UTC
收录当时偏移：UTC+02026年9月12日 13:01 UTC
- 发布
- 2026年9月11日 22:03
- 收录
- 2026年9月12日 13:01
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 正常
档位是按信源手工设定的编辑判断，不是逐条打分。
A man finds a magic lamp with a djinn inside. The djinn says "you can wish for anything you want. But whatever you wish for, your neighbor gets twice as much". After thinking for a while the man says "take out one of my eyes".
How are we supposed to trust that their version of alignment doesn't just mean "obey me, personally" and not use their massive leverage over everybody else?
Come to think of it, unaligned ASI has better odds of doing the right thing for everyone compared to a random person.
Nobody else wants you to succeed. Certainly not more than them. So how are we supposed to think people agree upon a definition of alignment that includes the well being of all of us and not just a few?