PickVidsWe Were Right! Real Inner Misalignment
推荐理由
There have been actual experiments in which AIs appeared to successfully learn their objective in training
说明
这是2021年由Robert Miles AI Safety 发布的一部讲解，主题归在AI safety, AGI & impact。它讲的是ai alignment这一支里的具体问题，而不是把整个领域扫一遍。适合已经能看懂基本术语、想继续往下走的人。内容上可以概括成：Robert Miles covers a paper where researchers ran real versions of his earlier mesa-optimizer thought experiments. They show inner misalignment, where an AI learns the wrong goal in training and misbehaves when deployed.这些读者给出的理由，核心是：Pointed to as evidence that AI misalignment is an empirically demonstrated problem, for people who dismiss it.讲法不绑某一条新闻，过几年仍然可以当材料看。从2023年到2023年，有1位不同的读者在1个不同月份把它推荐出来，其中明确叫别人去看的有0次。适合用一集把一个概念补上的人。抓住它用的那个图像或类比，比背公式重要。如果已经会了基本定义，就把注意力放在它怎么把ai alignment变成能跟着复述的例子。
视频信息
- 频道
- Robert Miles AI Safety
- 时长
- 11 分 47 秒
- 发布年份
- 2021
同主题更多
- What Happened to Horses Is Happening to Us
- Why Would AI Want to do Bad Things? Instrumental Convergence
- Veritasium: What Everyone Gets Wrong About AI and Learning – Derek Muller Explains
- Intelligence and Stupidity: The Orthogonality Thesis
- We're building a dystopia just to make people click on ads | Zeynep Tufekci
- Intro to AI Safety, Remastered