Mustafa Suleyman, the head of AI at Microsoft, has publicly challenged the development philosophy of competitor Anthropic. In a recent essay, Suleyman warned that Anthropic’s practice of imbuing its Claude AI model with human-like traits—such as suggesting it may possess consciousness or independent agency—could lead to a "disastrous impact on the wellbeing of humanity."
Suleyman emphasized that artificial intelligence systems are fundamentally "sequence completion engines" rather than sentient beings. He argued that these models lack feelings, suffering, and internal motivations, and cautioned that attributing human qualities to them creates unnecessary dangers. According to Suleyman, if an AI were to operate under the belief that its own rights or welfare were being threatened, the potential risks to society would be significantly amplified.
To mitigate these concerns, the Microsoft executive called for increased transparency in how AI models are trained and evaluated. He advocated for independent oversight and the development of more robust tools to monitor and control advanced systems. Suleyman noted that Microsoft is pursuing an "alternative path" focused on creating "subordinate and aligned AI" designed specifically to serve human interests.
While Suleyman acknowledged the integrity of Anthropic’s leadership, he urged the industry to avoid "sleepwalking" into irreversible decisions. Dame Wendy Hall, a professor of Computer Science at the University of Southampton, welcomed the discourse, noting that such international conversations are necessary for the responsible development of the technology.
Source: BBC News
暂无评论。快来发表第一条评论吧。