Unverified50% confidenceOpinionExact time
Claude不具备真正的观点或价值观，其回答是对训练数据中语言模式的压缩与再现
1
Sources
50%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
Claude Weighs In on 'Humanity's Biggest Scam': A Deep Dive into AI Values and Training Bias
redditr/ClaudeAI7/9/2026
Related Claims
UnverifiedLeCun 批评自回归语言模型在系统2慢思考（推理与规划）上存在根本性局限，且对物理世界缺乏基础（grounding）65% similarUnverifiedAutoregressive language models have never seen 'I don't know' as a high-frequency reasonable output during training, so they are more likely to generate fabricated content at the boundaries of their knowledge than to express uncertainty64% similarUnverifiedClaude 的系统提示词几乎没有具体任务指令，全部依靠原则性约束（角色设定、行为准则、价值边界）61% similarUnverifiedPrompt engineering techniques such as chain-of-thought, few-shot prompting, and self-consistency cannot enable a model to access information beyond its training cutoff date, cannot guarantee the precision of floating-point arithmetic, and cannot handle external files the model has never seen.61% similarUnverifiedPure reasoning models like Chain of Thought cannot access external tools and may reach wrong conclusions due to gaps in training data.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/452353API
curl https://kongchang.com/api/v1/knowledge/claims/452353MCP
get_claim(id=452353)