告诉 HN：不要将您的 AI 研究 IP 托付给大型科技公司（Bigco）的 AI 代理。
8 分•作者: botencat•3 个月前
我对于与 Claude [Code] 或 ChatGPT/Codex 等工具分享潜在的 AI 研究感到非常警惕。
我相信任何公司本质上都是一个“回形针最大化器”，它们会不择手段地赢得竞争。
AI 公司可以获取数百万与其直接竞争的 AI 研究人员和 AI 初创公司的知识产权。如果它们能够利用这些数据来扼杀竞争（无论是来自同一研究人员还是其他人的竞争），我相信它们最终会（如果还没有的话）使用这些数据，即使它们声称不会这样做。
它们不必公然窃取——它们可以仅仅在数据上进行训练，或者将“可疑”的聊天记录传递给人类检查员，而这些检查员最终可能会在自己的研究中受到“启发”。我们在 Fable 发布初期就看到了第一个（？）迹象，Anthropic 声称将降级关于“前沿 AI”（即任何与其竞争的产品）的模型响应。
从其他领域我们知道，例如 Uber 利用用户的乘车数据来扼杀竞争和监管。没有理由相信大型 AI 公司不会做同样的事情。
查看原文
I am very paranoid about sharing potential AI research with e.g. Claude [Code] or ChatGPT/Codex.<p>I believe that any company is essentially a paperclip optimizer that will do whatever it takes to win over competition.<p>AI companies have access to the IP of millions of AI researchers and AI startups who are in direct competition with them. If they can use this data to squash competition (either competition from the same researchers or from others), I believe that they will use it eventually (if not already), even if they say they won't.<p>They don't have to blatantly steal it - they can just train on it, or pass "suspicous" chats to human inspectors who might eventually be "inspired" by it in their own research. We saw the first (?) hint of this during the brief Fable release, with Anthropic declaring that they will downgrade model responses regarding "frontier AI" (i.e. anything that competes with them).<p>From other domains, we know for example that Uber used users' ride data to stiffle competition and regulation. There should be no reason to believe that Bigco AI companies won't do the same.