特朗普向Anthropic提出不可能的要求
阅读原文· garymarcus.substack.com特朗普要求Anthropic完成不可能的任务，暴露了生成式AI安全护栏的根本困境。早在2024年1月，Gary Marcus就指出任何护栏都难以在过于严格和过于宽松之间找到平衡。如今这一判断得到验证：基于next-token predictor的大语言模型本质上不适合安全控制。要么对LLM加以限制直至出现更好的技术，要么承受后果。问题并非Anthropic独有，而是整个生成式AI面临的挑战。
Breaking: Trump asks the impossible of Anthropic
Where do we go from here?
In January 2024, I warned that the politics and inadequacy of guardrails would become a central issue for our times.
It took longer than I expected, but well, here we are:
Where do we go from here? At least with respect to LLMs, the security experts are right. And the writing has been on the wall literally for years. As Katie Conrad and I wrote here in January 2024:
virtually any guardrail has to thread a needle between the Scylla of being too restrictive and Charybdis of being too permissive. None thus far have done this effectively.