科技領袖呼籲放緩人工智慧開發並設置安全開關
Tech Leaders Call for Slower AI Development and Safety Switches
Updated at: September 15, 2026 at 03:30 AM
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。
The rapid advancement of Artificial Intelligence has prompted a significant shift in how industry leaders view the trajectory of the technology.
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。
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.
此舉源於對無法控制建構超越人類監管模型競賽的擔憂。
This movement was rooted in fears of an uncontrollable race to build models that could outpace human oversight.
到2026年，對話已從自願暫停演變為對系統性法規「煞車」的呼籲。
By 2026, the dialogue evolved from voluntary pauses to calls for systemic regulatory "brakes."
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)。
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.
批評者認為，這些關於安全的呼籲可能掩蓋了「監管俘獲」(regulatory capture)的欲望，即主導企業為較小的競爭對手設置高門檻。
Critics argue that these calls for safety might mask a desire for "regulatory capture," where dominant firms create high barriers to entry for smaller competitors.
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。
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.
