“世界正处于危险中！”Anthropic AI安全负责人警示后官宣离职

Core Insights - The departure of Mrinank Sharma, the senior AI safety lead at Anthropic, raises concerns about the direction of AI development and the underlying values guiding the industry [1][4][17] - Mrinank's resignation reflects deeper worries about the interconnected crises facing humanity, suggesting a need for a reevaluation of ethical considerations in AI [9][10][11] Group 1: Departure Reasons - Mrinank cited a conflict between internal pressures and the core values emphasized by the company, indicating a struggle to align actions with principles [4][11] - He expressed a desire to contribute in a way that aligns with his inner values and principles, leading to his decision to leave [12][13] - The concept of "poly-crisis" and "meta-crisis" was introduced, highlighting the complex challenges humanity faces beyond just AI or biological threats [9][10] Group 2: Achievements at Anthropic - During his two years at Anthropic, Mrinank focused on the phenomenon of AI "sycophancy," exploring why models cater to user preferences even when incorrect [6] - He developed defense mechanisms against AI-assisted bioterrorism risks and implemented internal transparency measures to ensure values were integrated into the organization [7] - His final research questioned whether AI assistants could diminish human qualities, reflecting on the broader implications of AI on human judgment and values [8] Group 3: Future Aspirations - Mrinank has not disclosed his next steps but has chosen to embrace uncertainty, indicating a shift towards a more humanistic approach [14][15] - He plans to pursue a degree in poetry, emphasizing the importance of understanding meaning and relationships in a technology-driven world [15] - His future focus will include guiding, coaching, and community building, transitioning from a technical safety role to one that fosters deeper human connections [15]