TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

NIST Draft Guidance Inherently Hostile to Open-Source AI Models

26 点作者 giuliomagnifico9 个月前

3 条评论

lupire8 个月前
They forgot to say how
评论 #41522242 未加载
campuscodi9 个月前
That's why their draft documents, so you can make public comments. If the drafts are getting hostile, means most feedback has pushed them that way.
评论 #41520858 未加载
blackeyeblitzar8 个月前
In case people are wondering, these guidelines can end up being more like strict requirements in real life. For example, a partner company or customer company might require following such guidelines, even if they are just labeled “best practices”, as part of a contract with your company. Or maybe your company’s insurance agency requires it. Or maybe your auditing firm will refuse to sign off on something without implementing such guidelines. My point is, even something advertised as “best practices” can become more like “forced requirements”, in many different and surprising ways.<p>In the case of AI, you can see how a president’s executive order (<a href="https:&#x2F;&#x2F;www.whitehouse.gov&#x2F;briefing-room&#x2F;presidential-actions&#x2F;2023&#x2F;10&#x2F;30&#x2F;executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.whitehouse.gov&#x2F;briefing-room&#x2F;presidential-action...</a>), which is what pushed NIST to create this draft, can ultimately force private companies to adhere to a president’s whim. This can be especially dangerous when you consider that these guidelines discuss misinformation, disinformation, and “hateful” content, as this feels a lot like violating first amendment rights through multiple steps using the flimsy claim that such guidelines are merely voluntary.