About

关于我

Hi，我是家泽，一个AI 对齐研究工程师。

我对AI Safety相关的工作感兴趣，不只是提升现有LLM的能力，更是对于如何对齐一个人类同样水平的AI。

我对斯多葛学派比较感兴趣，也在阅读相关的书籍。

日常生活中，一方面在研究如何提升现有Sota模型的能力，另外一方面在研究如何提升自己。

除此之外，我还是一个小姑娘的爸爸，观察陪伴孩子的成长比训练模型更有意思。

关于博客

在过去的一年里面，我读了很多非常棒的博客文章，给了我很多启发。我也希望能写一些感兴趣的东西来记录，同时也能对我或者其他读者有所启发。

博客名字来源于：我们发现bug的过程，源自于我们发现了一些不对劲(seems wrong)的地方。同时，找到bug，也意味着离解决问题更进一步。

其他的一点小想法：在做模型预训练的时候，天然会发现中文互联网的高质量语料比英文高质量语料少太多了。希望能做一些微小的贡献，提供一些略微超过平均的语料。

---

About Me

Hi, this is Jiaze, an AI Alignment Research Engineer.

I am interested in AI Safety, not only in improving the capabilities of LLMs, but more importantly in how to align a human-level AI with humans.

I am quite interested in the views of the Stoicism and am currently reading related philosophical books.

Recently, what I have been doing is a combination of two things: on the one hand, I am using the views of Stoicism to improve myself; on the other hand, my main research focuses on how LLMs can improve their capabilities through self-improvement.

Besides, I am also the father of a little girl, and observing and accompanying my child's growth is more interesting than training models.

About the blog

In the past year, I have read many excellent blogs written by inspiring individuals. I also hope to document things that interest me to see if I can inspire myself or others who read my blog.

The name of the blog comes from: when we find bugs, we always say that something seems wrong here. Finding what’s wrong seems to be the beginning of success.

I hope everyone can notice things that seems wrong in one’s life, then follow the clues to find the root causes, fix the problems, and achieve self-improvement.

Another small thought: When doing model pre-training, I naturally found that the high-quality corpus of Chinese is much less than that of English. I hope to make some small contribution and provide some corpus that is slightly above average.