Posts

Showing posts from January, 2024

视频推荐:文化进步与保守坐标图

Image
推荐一个mhyyyy老师的视频:《文化进步与保守坐标图》。 https://www.youtube.com/watch?v=DiDye3-YJ24 文化进步与保守是除了经济左派/右派,政治权威/自由,之外的另一个重要的政治光谱。本视频列举了所有常见的进步/保守观点上的意见,并依序排列,量化它们。 注意:保守不一定意味着差,而进步也不一定意味着好,这只是2个没有褒贬的客观的词汇而已。 除此之外,本视频还介绍了以下几点: 首先,进步/保守是相对的。比如说,支持妇女投票权在19世纪,乃至20世纪的初期,都是相当进步的。但是目前几乎没有哪个国家会认为妇女和男性的投票权应该有所不同。此外,lgbt在我的小时候实在是一个几乎没人会关注的话题,而现在,大家,尤其是女生,都已经觉得这件事情可以拿上台面来讨论了。而在欧美的一些国家,甚至同性婚姻已经合法了。 其次,中国人平均相当保守,大约相当于欧美的右翼。而中国大学生的平均保守/进步程度大约相当于世界平均水平。本视频有一个思路特别好,它把“中国人”的文化轴定量化,去媚化了。就是说,中国没有什么特殊的地方,中国的很多特殊国情也和世界上其他地方的一样。免得有人一说起lgbt,或者自由民主,就说中国的国情有别,中国人不能用西方的方法去量化。 此外,年轻人中,男性往往比女性保守。我认为是因为女权主义确实带给年轻女性更少的伦理束缚,更自由/多元的选择。总而言之,女性通过支持更进步的女权主义,而获得更高的社会地位。相比之下,有不少男性选择回归传统,推动社会向保守方向发展,来维持自己的高地位。屁股不一定决定脑袋,但是大概率。 按照这个指标,我本人的进步坐标值大约是8.5,介于进步8-进步9之间,我的很多博文是站在相当进步的角度上去写的。

统计局在报道全国固定资产投资时数据造假 - The Bureau of Statistics falsified data in their report on national fixed-asset investment.

Image
由于本文可能稍有敏感,这里仅展示英文版。  The National Bureau of Statistics of China made errors in the growth rates of national fixed-asset investment for the years 2023, 2020, and 2019 in its report. This serves as proof that the government sometimes may not be trustworthy. Image Source: torontobigface https://twitter.com/torontobigface/status/1752182533582954558?s=46 Upon verification, each piece of information turned out to be authentic. --- Furthermore, this lends credence to Hideo Tamura's suspicions that China's GDP figures may be fabricated.  

使用贝叶斯方法分析汪文斌“新闻自由”言论 - Using Bayesian methods to analyze Wang Wenbin's remarks on "freedom of the press."

Image
由于本文可能稍有敏感,这里仅展示英文版。 On January 2, 2024, Chinese Foreign Ministry Spokesperson Wang Wenbin stated, "The so-called 'freedom of the press' is merely a tool used by certain countries to attack and slander other nations." This statement has sparked controversy and is considered extreme by many. Let's analyze its credibility using the Bayesian approach. Introduction to the Principle: A rational person's method of judging the truth or falsehood of a matter can be viewed as a Bayesian model. When someone says something that aligns with our understanding of an issue, we tend to strengthen our belief in that issue and the credibility of the person increases in our eyes. Conversely, when someone says something that contradicts our understanding, two scenarios may occur: If the person has good credibility and our knowledge of the issue is not extensive, we tend to believe the person and revise our understanding of the issue. If the person has poor credibility and our...

推荐一个开源大语言模型:Mixtral-8x7b - Recommand a Large Language Model: Mixtral-8x7b

Image
大语言模型(LLM)是用来生成文本的AI模型。举一个例子:ChatGPT就是一个闭源的大语言模型。 Large Language Models (LLMs) are AI models designed for text generation. For example, ChatGPT is a closed-source large language model. 相比于闭源的模型,开源的模型免费,并且没有审核限制,可以随意使用。但一般的开源大语言模型的水平往往很差。 Mixtral-8x7b是一个由Mistral AI公司发布的开源的大语言模型,截至目前,它的“水平”是开源模型中最高的,在大多数测试中其性能与gpt3.5类似。 In contrast to closed-source models, open-source models are free and lack moderation restrictions, allowing unrestricted use. However, the quality of most open-source LLMs is generally subpar. Mixtral-8x7b is an open-source large language model released by Mistral AI. As of now, it is the highest-performing open-source model, with capabilities similar to GPT-3.5 in most tests. Mixtral-8x7b可以在这里使用。点击链接,在右下角选择模型处选择Mixtral-8x7b-instruct,就可以开始与它对话了。 Mixtral-8x7b can be accessed here. By clicking the link, selecting 'Mixtral-8x7b-instruct' at the bottom right, you can start conversing with it. https://labs.perplexity.ai/ 此外,也可以在Google Colab部署Mixtral-8x7b。 Additionally,...

热门博文 - Popular posts