马斯克突然赞成AI减速：顶级大佬为何集体踩刹车？

真正让大佬踩刹车的，可能不是AI不够强，而是AI强到开始超出预期了

Alan 每次都把事情分析得很透彻👍

谢谢🙏

结果把大盘刹熄火了

假摔？

这么说的话确实马斯克的提议是最优解

老马有一套

Agents’ “just do it” spirit makes the risks much more unknown. It’s beyond LLMs giving wrong answers, now it’s about real actions to follow any prompts… Some incidents could go really big, which could mean lawsuits, fines, regulatory investigations & possible congressional hearings…No big organization would risk it all, especially when things could simply happen with those highly capable agents. The guardrails need to be developed to test the advanced models and agents, so the risks can be minimized prior to releasing to the grand public… Not sure which parties are truly capable of doing this work, but the results carry a lot of responsibility. Eventually, it could be something like the OCC for banks, a new federal organization responsible for AI regulation, enforcement, and regulatory fines. Don’t hate me, but without proper controls, individuals or other entities using AI products could be negatively impacted without enough guidance or protection.

agree

要真正看到“牆”，才可能定規則。 這裡所說的牆是資源與物理上限 。沒有錢燒了 。算力與模型沒能再快再玩了 否則都是喊喊而已。

interesting shift, maybe they see the risks clearly now.

partially