主题：【原创】机器学习的基础是什么？(0) -- 看树的老鼠

1.

"Moravec的悖论", a great one. thanks to "Fuhrer".

http://www.ccthere.com/article/3839163

and "Fuhrer" may not be happy with my 阶级斗争 interpretation, but 阶级斗争 is what my model does(:), I like prof. Frank Wilczek models, but I am not really in QCD business.

most of today's AI etc is almost all about 阶级斗争, but we have to have non-阶级斗争-interested physicists such as prof. Frank Wilczek provide us with models so we can play 阶级斗争 game.

or we go back to the "hot war" form of 阶级斗争, mostly leveraging on Newton mechanical physics.

2.

达尔文世界的 business model

达尔文世界=耗散系统的物理实现, 人类文明系统必须加速度吐故纳新,加速度进步.

人类文明没有共产主义, 只有残酷进步,进步通过系统内部残酷竞争实现,吐故纳新,最小作用原理

heatbath, 虚功原理, etc: 热力学系统本质上是内能过程, 时间箭头, "进步".

达尔文世界的社会政治经济系统逻辑=无限追求创新增长, 或者死亡; 创新增长, 信息, to be rewarded by system with huge premium, everything and everybody else to be priced at steep discount. 个人, 公司, 国家, 无一列外.

3.

阶级斗争需要巨大的计算资源

"高层次的推理只需要很少的计算", such as accounting:

"但低级别的感觉运动技能需要巨大的计算资源", such as

"本次中石化混改",

"这次表面看问题不大实际问题可能更大" 花21 陈王奋起挥黄钺

http://www.ccthere.com/article/4051447

"Moravec的矛盾是人工智能和机器人技术研究人员发现，传统的假设相反，高层次的推理只需要很少的计算，但低级别的感觉运动技能需要巨大的计算资源。汉斯·莫拉维克，罗德尼·布鲁克斯，马文·明斯基在20世纪80年代明确提出的原则。莫拉维克写道，“这是比较容易使计算机具有成人级的性能智力测验或玩跳棋，很难或根本不可能给他们一岁的时候，它涉及到感知和流动性的技能。”

4.

阶级斗争 is more of ""强相互作用" type of physics, where you have strong coupling/interactions among system's degree of freedoms, simply speaking.

witten1:

"本质上你还是得把配分函数弄出来才能得到一些启动那个微分方程的量，而这对于full interaction的体系，这个配分函数要是能有效算出来，这工作的难度和分量我想你是明白的－－breakthrough。。。"

witten1:【原创】量子生物学I 摘要和前言分页第3页 - 西西河

www.cchere.com/topic/3830274/3

配分函数 for a full scale non-linear system? game over.

that is why 量子生物学/量子分子学 is hard.

but we can't afford game over in physics, so we put in a "cut-off" in terms of energy level, because coupling/interaction is associated with high energy level: if you don't know how to handle it, you cut it out from your model.

besides, "低能区的物理对截断的形式并不敏感".

with non-生物学 physics, we can do it, we largely live in an stable and near equilibrium physical world, where energy level is limited

but you cannot do it with 生物学 (including 基因食品), 阶级斗争, that is one of reasons why 阶级斗争 is very hard to model.　

example: I was googling 刘川, but google gives me 刘诗诗, understandable, who cares about 刘川's cut off?

刘诗诗 is gorgeous, making 阶级斗争 even more non-linear(:)?

wherever she appears, we need to somehow update 配分函数, a must.

-----

and by the way, 清华大学材料系教授 韦丹, his wife, 韦丹 固体物理讲义 is pretty good.

about cut-off/能动量截断

"刘川的量子场论讲义：对于一个有相互作用的量子场论来说，由于量子涨落的“虚过程”可以在任意能动量发生，因此，如果理论不存在某种能动量截断，那么由于长是互相作用的，一个低能动量的模式就可以通过虚过程与无穷高能动量的模式发生相互作用，这就造成了场论中的“紫外发散”（即计算得到的结果是无穷大）。而如果量子场论中存在一个物理的能动量截断，它就可以保证所有的量都不发散。这个截断的具体形式其实对于远低于截断能标的物理来说并不重要，因为低能区的物理对截断的形式并不敏感。重要的是，这个截断是存在的，而且它是相互作用量子场论不可或缺的组成部分"