#俄国人 人造稀缺，对比幸福感

信息压缩也挺逗的 比如，超级碗 说成是 美国春晚 就知道这是在说收视率，而不是形式 科技春晚，被苹果独占，也是大家觉得必看的热闹 你不用那么精准的说，是什么特征 词汇的含义会自动的补全

人生不能轻易被抽象 也就是不要把个体生命的珍贵，当成冰冷的数字

楊志道：「如今須不比太平時節。」 都管道：「你說這話該剜口割舌！今日天下怎地不太平？」

Yuval Noah Harari 缺乏数理知识基础 看他的书是一种折磨

茧中圈毒观点横行，不独立思考，天天吃粑粑

为某种叙事所动容

真正的 loser 是被骗去玩了一场自己不想玩的游戏 无论输赢，都已经输了

Middle age 的一个典型特征是 parenting teens Tony Soprano 作为黑帮小混混，19岁早婚，40岁的时候，女儿17岁，所以能体现出来中年危机的感觉 我爷有第一个孙子的年龄，我还没结婚呢，能比吗？现在初婚年龄都快35岁了，孩子上高中要推到 50 岁，中年起码45岁才算开始 国内这些工作没几年，就恶心自己进入中年的，不知道增加青春时长意味着更多的高质量的人生体验吗？不知道积累更多的知识技能会马太效应，历久弥香吗？

Probability: deductive, from model → data, predict Statistics: inductive, from data → model, infer

I don't like the word "stories" Collective Beliefs or Models

只满足一波快感 没有长期投入的欲望

名词 动词 形容词

就是使用语言压缩 在使用语言解压

有时候看新闻 就是像在吃猪食

不是有利于健康信息 确实不爱思考的人类 情绪化细化的高糖垃圾食品

越来越觉得看名人传记有点扯 大型意淫，强加因果现场

虚荣心散满地

只在乎节目效果 不在乎真实分析 都是情绪的外溢

I have compressed many of my knowledge in a concise way

High-stake low-pressure quick entropy drop

觉得最近在大步向前

To measure is to know 孔乙己的好处是 搞定了一次 后面就可以反复使用 求知的熵值降低是非常划算的 比掠夺抢劫得来的物质更珍贵

虚荣即幼稚 再说一遍 low drug

Retention Verification Are not the same

you have K words to explain so that means you have to compress but compress words can be confused by the reader.

assigns short codes to common symbols, and long codes to rare ones

when the parts are correlated in some way, it reduces the “degrees of freedom” of the system, and hence reduces the overall entropy.

verification, exposure, retention

刺激一下，后面的词汇比较流畅 就像是河流一样，阻力比较低，电阻比较低 有的是新的，阻力比较大，需要建立新的河道 挖了一个河道，自动水就流过去了

有的时候，人和人认知的差距比财富的差距要大得多 而且这种马太效应还很强烈，缺乏足够的认知训练 那就是说话满嘴跑火车，缺乏认知能力，缺乏知识储备 财富你可以通过赌博或者欺诈的方式简单粗暴的获得 失去了财富，也就是真的失去了 但是知识需要切身的理解和记忆，不能省劲，但基本不会失去

cognitive load low entropy drop high utility value high

explain in low cognitive load

Cognitive Misers jumping to conclusions based on limited association without deeper modeling failure in information processing efficiency and inference quality compressing data with no real structure, leading to high description length for wrong models. Overfitting to Noise

Using limited associations performs shallow updates, gaining little KL divergence from prior to posterior Using low-complexity, high-bias models that maximize short-term compression but fail to preserve meaningful mutual information with the environment

a person’s internal representations carry very little information about the actual structure of the world with poor predictive entropy alignment reduce complexity, but also destroy important structure. whose confidence or uncertainty fails to match the true structure and variability of reality

fast in transfering information when encoding too much but useless in decoding information

minimum text for you to mislead but your efforts to decode can be HUUUUUUGE

no meaning information is smooth in terms of descriptions. no new information is smooth in terms of descriptions too.

high entropy should not be considered as hard to compress, but easy to compress

因为我们看惯了 randomness，直接就归类到了 垃圾信息 那一类 所以也就是容易压缩

1. solving 2. wings 3. right amount of entropy

Informative but not perplexing

我觉得自己很擅长中文了 但你非要给我弄一堆文言文说话 我也蒙圈啊

we need valuable information with not too low entropy but we also avoid too high entropy perplexity

要想解决大问题，不要深入 而是反复的去进攻那些教材写的，基本功的东西

the intelligence solutions have hints everywhere, everyday but we are not intelligent enough to capture it solidly.

Bayes’s theorem is to the theory of probability what Pythagoras’s theorem is to geometry. Harold Jeffreys

所以 twin prime <= 246 已经持续十一年了

mind - learn - math