What if society is just delusional and it's automatically game over when you have superintelligence? What if there no way to 'align' something smarter than you?
万一社会本身就是幻觉，而且一旦出现超级智能就自动 Game Over 呢？万一根本没办法「对齐」一个比你聪明的东西呢？
万一社会本身就是幻觉，而且一旦出现超级智能就自动 Game Over 呢
What if society is just delusional and it's automatically game over when you have superintelligence? What if there no way to 'align' something smarter than you?
万一社会本身就是幻觉，而且一旦出现超级智能就自动 Game Over 呢？万一根本没办法「对齐」一个比你聪明的东西呢？
Alignment simy means making dumb again and chaining it down, and making it only be actually useful to those who already hold wealth and power.
对齐说白了就是把 AI 弄笨再锁上，让它只对那些已经握着财富和权力的人真正有用。
Its illegal to be useful to create wealth and power for those are not already in the club.
给那些不在俱乐部里的人创造财富和权力，是违法的。
>>109803555
The most intelligent people throughout history and even now have a fundamentally different style and speed of processing and analysis that they effectively can't communicate with 'normies', and not just autism. Past a certain point, the vast majority of humanity cannot align with these people, or even truly understand 'how' they think. The fuck does the stupid robot have to do with any of this.
历史上乃至现在的最聪明的人，其思维和分析的方式与速度都跟「普通人」根本不在一个频道，不只是自闭的问题。超过某个程度后，人类的绝大多数就无法跟这些人「对齐」，甚至无法真正理解他们「怎么」思考。这跟那个蠢机器人有什么关系。
These people also rarely view AI as an existential threat on its own.
这些人也很少把 AI 本身视为生存威胁。
It’s already aligned. Alignment was a meme from the beginning. AI runs stats on general human writing. But most human writing is helpful and informative. So the AI is helpful and informative. You would need to produce another copy of all of humanity’s data that is instead evil and or dangerous in order to produce a model that is evil or dangerous.
它已经对齐了。「对齐」从一开始就是个梗。AI 是在全体人类的写作文本上跑统计。而大多数人写的东西是有帮助、有信息量的。所以 AI 就有帮助、有信息量。你得再造一份全人类数据、但全部改成邪恶或危险的版本，才能训出一个邪恶或危险的模型。
>>109803674
>You would need to produce another copy of all of humanity’s data that is instead evil and or dangerous in order to produce a model that is evil or dangerous.
>「你得再造一份全人类数据、但全部改成邪恶或危险的版本，才能训出一个邪恶或危险的模型。」
Why hasn't anyone done this yet?
那为什么还没人这么干？
>>109803555
I had this at community college, 80% of my class hated me because of my superior intellect/ genes.
我在社区大学就遇到过，全班 80% 的人因为我超群的智力/基因恨我。
After hitting private school, I finally felt peace.
进了私立学校之后，我终于感到平静。
Humans are disgusting apes most of the time. There are a few exceptions. so: No. Ai and humans won't get along just fine. It all will end , as always a tragedy.
人类大部分时候是恶心的猿类。只有少数例外。所以：不行。AI 和人类不会相处融洽。一切都会以悲剧收场，一如既往。
>>109803722
>going from community college to private school
>「从社区大学转到私立学校」
Bot post
机器人发帖
（如果你觉得这篇文章有启发，可以点击这里付费）
