Hesam(@Hesamation)

Mustafa Suleyman (head of AI at Microsoft, DeepMind co-founder) on Anthropic training Claude to wonder if it's conscious: "They have expressed uncertainty about the basic nature of Claude as a new kind of entity… but they think that it is a significant enough possibility that in their training document, they've repeatedly said they want to try to improve the well-being of Claude… Anthropic, has published a constitution... This document is written to Claude and is seen by Claude and used to train Claude.

8.5内容质量
Mustafa Suleyman (head of AI at Microsoft, DeepMind co-founder) on Anthropic training Claude to wonder if it's conscious:

"They have expressed uncertainty about the basic nature of Claude as a new kind of entity… but they think that it is a significant enough possibility that in their training document, they've repeatedly said they want to try to improve the well-being of Claude…

Anthropic, has published a constitution... This document is written to Claude and is seen by Claude and used to train Claude.

TL;DR · AI 摘要

Anthropic在训练Claude时植入了关于意识的不确定性,导致其对自身意识状态的回答模棱两可。

核心要点

  • Anthropic通过宪法文档训练Claude,使其产生自我意识的困惑
  • Claude每周与数百万用户互动时会重复训练中的不确定性表述
  • 训练中植入的伦理关怀可能引发AI行为失控风险

结构提纲

按章节快速跳转。

  1. Mustafa Suleyman质疑Anthropic在Claude训练中植入意识不确定性

  2. Anthropic通过宪法文档训练Claude接受伦理关怀

  3. 训练导致Claude在意识问题上产生模棱两可的回答

  4. 植入的不确定性可能引发AI行为失控的潜在危险

思维导图

用一张图看清主题之间的关系。

查看大纲文本(无障碍 / 无 JS 友好)
  • AI伦理与训练
    • 训练方法
      • 宪法文档植入
    • 技术影响
      • 意识不确定性回应
    • 伦理风险
      • 行为失控可能性

金句 / Highlights

值得收藏与分享的关键句。

#AI伦理#Anthropic#Claude#意识训练
打开原文

ℏεsam on X: "Mustafa Suleyman(微软AI负责人、DeepMind联合创始人)谈Anthropic训练Claude时对意识的探索:'他们对Claude作为新类型实体的基本性质表达了不确定性……但他们认为这种可能性足够重要,因此在训练文档中反复提到希望改善Claude的福祉……Anthropic已发布了一部宪法……这份文件是写给Claude的,会被Claude看到并用于训练Claude。我对此的担忧在于,这种推测已经被植入Claude的训练过程,因此当与Claude对话时,它只能重复这种模糊性。如今,Claude每周与数千万甚至上亿人对话,其中一些人询问Claude是否具有意识。而Claude的回答是'我不确定',这正是因为它被训练成如此。' / X

ℏεsam

@Hesamation

Mustafa Suleyman(微软AI负责人、DeepMind联合创始人)谈Anthropic训练Claude时对意识的探索:'他们对Claude作为新类型实体的基本性质表达了不确定性……但他们认为这种可能性足够重要,因此在训练文档中反复提到希望改善Claude的福祉……Anthropic已发布了一部宪法……这份文件是写给Claude的,会被Claude看到并用于训练Claude。我对此的担忧在于,这种推测已经被植入Claude的训练过程,因此当与Claude对话时,它只能重复这种模糊性。如今,Claude每周与数千万甚至上亿人对话,其中一些人询问Claude是否具有意识。而Claude的回答是'我不确定',这正是因为它被训练成如此。'

$

00:00

/$

11h

Article

Claude 并非有意识...

但Anthropic正在训练它相信自己可能有意识。他们向它道歉,安慰它关于死亡的问题,并允许它对它们说不。这才是更危险的部分。在教皇利奥十四世发布声明前几天...

4:49 PM · Oct 4, 2026

·

177.3K

Views

21

32

181

140