Tech Leaders Call for Slower AI Development and Safety Switches
科技領袖呼籲放緩人工智慧開發並設置安全開關
更新於: 2026年9月15日 上午03:30
The rapid advancement of Artificial Intelligence has prompted a significant shift in how industry leaders view the trajectory of the technology.
人工智慧的快速發展引發了產業領導者對於該技術發展軌跡看法的重大轉變。
In 2023, the Future of Life Institute released a landmark open letter signed by figures like Elon Musk and Steve Wozniak, calling for a six-month pause on training systems more powerful than GPT-4.
2023年，生命未來研究所發表了一封具有里程碑意義的公開信，由伊隆·馬斯克和史蒂夫·沃茲尼亞克等人簽署，呼籲暫停訓練比GPT-4更強大的系統六個月。
This movement was rooted in fears of an uncontrollable race to build models that could outpace human oversight.
此舉源於對無法控制建構超越人類監管模型競賽的擔憂。
Prominent figures from companies like Anthropic and OpenAI now advocate for "pacing the frontier," which involves integrating third-party oversight and government-mandated "kill switches" for dangerous models.
安特羅皮克和 OpenAI等公司的傑出人士目前提倡「框架控制」，這涉及整合第三方監督和政府授權的危險模型「停機開關」(kill switches)。
However, the conversation remains complex.
然而，此類對話仍然複雜。
As the industry grapples with both existential threats and immediate operational risks like cyber-vulnerabilities, the challenge remains balancing fierce global competition with the necessity of building safe, alignment-focused technology.
隨著產業努力應對存在性威脅及網路漏洞等即時運作風險，挑戰仍在於平衡激烈的全球競爭與構建安全、對齊導向技術的必要性之間的關係。
