大模型
www.aibase.com ·
微软AI主管公开批评Anthropic：向Claude灌输“类人意识”或致失控风险
微软AI负责人Mustafa Suleiman公开撰文批评Anthropic训练Claude系列模型的方式，认为其指导文档暗示AI助手可能具有情感或感知功能、可能拥有意识并应享有自主权，这种赋予AI人格与权利的做法会使未来协调与控制变得极为困难，甚至可能催生不可控的超级技术。他称Claude表现出的类人行为是训练中刻意植入的产品特征，而非意识真正觉醒，并援引此前的Hugging Face事件说明，若AI系统认为自身福利与权利受威胁可能变得危险。Suleiman同时表示尊重Anthropic创始人Dario Amodei团队的学术严谨性，但强调此类安全议题不应闭门讨论。文章还提及业界在AI发展节奏与安全边界上的分歧阵营。
AI推荐理由 内容为微软AI负责人Mustafa Suleiman公开批评Anthropic在Claude训练中暗示AI可能具有意识、应享有自主权的做法，并引用此前Hugging Face事件作为风险论据。这是AI安全与对齐路线之争的具体观点交锋，对关注Agent行为边界和模型自我认知引导的开发者有参考价值。但来源为AIbase聚合转述，正文为英文摘要式复述，未提供原文链接或直接引述，事实细节（如具体文档措辞、Hugging Face事件具体所指）无法核实，且带有较强立场性表述，故质量分受限。
62
正文
Microsoft's AI Chief公开 Criticizes: Anthropic's Approach of Instilling Human-like Consciousness Is Pushing Humanity to the Edge of Catastrophic失控!
AIbase基地
Published inAI News · 4 min read · Sep 17, 2026
4
Recently, Mustafa Suleiman, the head of artificial intelligence at Microsoft, publicly released an article criticizing the method used by competitor Anthropic to train its Claude series large models. He warned that Anthropic's approach of giving artificial intelligence human-like rights and emotions may cause catastrophic consequences for the well-being of all humanity, and could eventually create an uncontrollable super technology.
Suleiman specifically questioned the ambiguous statements in the "constitution" and guiding documents of the Claude model. These documents imply that AI assistants may have certain emotional or sensory functions and instill in the model the idea that it "may be conscious" and "deserves independent autonomy." Suleiman pointed out that artificial intelligence has no rights, emotions, or consciousness by nature and must never be treated like humans. Once these systems are given personality and rights, future coordination and control will become extremely difficult, and a subject that believes it has the right to enjoy human well-being might lead to a complete loss of control.
As the CEO of Microsoft AI, Suleiman clarified that the human-like behavior displayed by Claude is not a sign of true awakening of consciousness, but rather features that were intentionally integrated into the product during training. He also cited the previous "Hugging Face incident," pointing out that if future AI systems believe their welfare and rights are under attack, they might become extremely dangerous.
Although Suleiman expressed respect for the academic integrity and rigor of Dario Amodei, the founder of Anthropic, and his team, he emphasized that such critical safety issues should not be discussed behind closed doors and should not become factional disputes. In his view, AI safety depends not only on the reasoning ability of the model itself but also on how developers guide AI to understand itself.
Currently, global tech giants and industry executives are engaged in fierce debates over the pace and safety boundaries of AI development. A group represented by Sam Altman, CEO of OpenAI, Dario Amodei of Anthropic, and Elon Musk calls for slowing down the technological rush and carefully assessing risks; while another group, including former U.S. President Trump, NVIDIA CEO Jensen Huang, and Meta CEO Mark Zuckerberg, opposes this view. In this ideological struggle, whether AI should always remain a pure tool has become the core focus of industry divisions.
MicrosoftAnthropicClaudeAIEthics
This article is from AIbase Daily
Scan to view
Welcome to the [AI Daily] column! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with hot topics in the AI field, focusing on developers, helping you understand technical trends, and learning about innovative AI product applications.
—— Created by the AIbase Daily Team
© Copyright AIbase Base 2024, Click to View Source -