I don’t know why I would choose some anime wallpaper as the theme for it, I really don’t know…
有时候想要写作的情感浴噴而出，但是如果沒有及時記錄，便會忘記。
你看，这次我就忘记了。
Last night at the most frightening moment worring about my mom, I dreamt of Moo Baan Dek, and all the children. I hugged them all, crying. I called their name, and asked where Mo is. I cannot understate the profound influence of Moo Baan Dek to me. It is one complete satisfaction to my broken childhood. I was healed, and I feel warmth. I stopped crying.
Thank you, MBD.
每次看到一个人弹着吉他清唱，然后轻声向上飞扬时，不知道为什么，我就想哭。每次看到一个人在阳光下奔跑，我就想哭。
我会永远为曼谷摩的上的人们祈祷。快捷，莽撞，精确，驾驶的艺术，风声，手心冷汗。我最喜欢花一半价钱打摩托车。东南亚湿热的风带走汗水，思绪在焦油后面粘稠，鱼在水道中穿行。永远车水之上的生计，在绿林灯火中吹出一朵热气迷狂的声浪。
然后他们低头致意，向着每个谦让的过客。Wai. Krub.
Observations of various, maybe different, complex systems
The expression level of a protein
牛顿从苹果的掉落中总结出了万有引力定律，我们能从观察蚂蚁和星群中看多远？
The good reason why the study of Mechanistic Interpretation, as in AI alignment, is that it provide a fast feedback loop and demonstrate core research mindset.
Recently interested in ML Alignment, more narrowly machanistic interpretability.
You are welcome to join me. Email: nathan16@g.ucla.edu
Bruin’s philosophy club is fun
Anthro, or archeology, has many innovative methods that is fundamental to study any things.
The crisis of self-identification let people learn I am in one of those crisis right now
Inavoidable in human nature, perhaps the chill days are the best days, and the most brillant.
Fantastic backpacking trip for the last weekend. Feels like found my group of people. It really turned out to be very hardcore. It was kinda raining, while we were determined to reach a waterfall. We barefooted trailed to a high waterfall sources. It took us around an hour to be there.
Went to the philosophy club. Normally I would be very active regarding to those topics, today I wasn’t really active. Some solitude is just can’t be resolved, and maybe finding a emotional and intellectual origince is just epheminal.
The journey is already worthwhile before I step my feet on the Thailand. The hardest part is to really make sure you are not in danger or getting scammed by a person you never met, and by an organization nobody around you have never came across with.
Now I am in Moo Baan Dek, kanchanaburi, thailand. Lovely kids, astounding school. beloved teachers, family.
The research field in mech corp give me a feeling of indecisiveness. Anthropic transformer threads and the LessWrong are two great places to get started. A few older papers comes from Distill. There is too much blogs to see, very busy.
Today I cleaned up a wasted room at the corner of the village, and moved in some old but excellent quality music instruments in. Journey just began.
Plan to focus on a few field during the summer: Computational complexity, topology, and computational neuroscience.
The script is finally working!
Teaching record: First week was watching movie: Dora the explore, the big red dog, and then combined with a lot of writing on the whiteboard. Writing on the white board can be very fun and engaging. Tell the kids to come up and draw pictures plus the English words, and followed by pasaThai. There isn’t a topic for us. Second week was pretty much the same. Third week I printed out some worksheet about body parts, and then did Simon says, and some songs and dancing. I feel like music/culture would be a great topic. It’s important to note that we are not here to be an ambassador of a very specific culture that we thought would be ahead of everyone else. We are here to promote knowledge and equality of cultures internationally. Thus it’s important for the children to respect their own culture, and then understand other cultures around the world, including Asia, pacific, Eastern, western… I also made use of many music instruments, drums and harps as they are easy to play. The class is very engaging!
Fourth week: Middle school lessons: First class: Making 3-fold name tags with names in the middle columns and favorite Animal, movie, music, food in the other columns. Then present basic self-introduction using the name tags. https://bogglesworldesl.com/lessons/Nov2001.html : that teaches ability to ask questions in English about words and sentences they don’t understand. English worksheets provided by teacher
Second + Third class: Roblox game development with AI
familiar with the game development interface AI included. Students learns quickly. Ask them now they will still remember how to use it. Mostly case-by-case demonstrations
introductions on operations of Roblox as a company models of profits and advertisement for
Fourth class: Nash Equilibrium and Group Theory in Math Nash Equilibrium for the importance of team work. Recognizing the similarities/differences in tic-tac-toe; Chess; prisoner’s dilemma; and then rock-paper-scissors. Group Theory in Math for a abstract way to understand patterns and symmetries.
Fifth class: Continuing Group Theory. Watching 3-blue-1-brown Youtube math video. Pal showed high interests.
Sixth class: Farewell. Teaching kids to write emails to me. Ben failed to register a google account.
Moo Baan Dek School Management notes and first draft:
The school don’t have any curriculum for common subjects. Each teacher decides on what to teach and design worksheets on a class-to-class basis.
There is no group chat for teachers. Sometimes teachers don’t know the schedule of the school or the events example: one teacher took some kids to see the doctor.
Children rarely read books.
Kids likes tictok
7/14 向伟大的人学习 He is the son of Mel Awl. He looks like Gandi, with the glasses on. He showed me the piles of documents about education for children with special need. All the note takings… He truly run the school. He sits outside with his computer, under the tree. He said he has to work outside even when it’s hot, because he has to see all the kids, and the kids can come to him whenever they like. 村一生 我很少参与种地，但是孩子们每天都做，我起床之前他们已经开始打扫屋子和周围，下午还会去花园以及农田和鸡舍。有些时候他们会在屋子里编制篮子或者做木工，绘画。在这个村子里生活，事实上我很少知道其他孩子在干什么，在哪里，但是什么都不知道也能生活下去。不问就不知，但即使妈妈们也不知道全部。外面我们也从来不会知晓所有事情，但有些事情已经被安排的严丝合缝，有些则是和你毫无关联。不论，我都从来没有被赋予过“知晓”的权利。但在这里，我会察觉到，询问任何事情，允许，不奇。没有网络，没有群聊，虽然我建议过妈妈。我们称呼大人/老师/监护者/职工 “Mel” 意为 “妈妈”。因为他们/她们教课，陪着孩子玩耍，夜晚也孩子们共居在零零散散的木头屋子里。妈妈全天，全周都在村子里，孩子们也是。大家有些时候会出去，看医生，参加活动，或者单纯游玩。但最近的县城开车也要半小时。全国每天中午都有捐赠团队带着他们的饭菜，物资进来，这就是这个村庄/学校：หมู่บ้านเด็ก “Moo Baan Dek” 赖以运营的一部分。每个人有一个大碗和一个勺子叉子，我也是。我们就用这个打饭，吃完之后洗干净碗筷。村里的事务分配的井井有条，食堂以洗碗池小孩子都一块打扫。吃剩的饭菜倒在一个桶里，要么被狗打翻吃净了，要么拿去decompose. 我写作之时大家都去上课了，我恰巧有空。这也是一天中最凉，最安静的时间，院子里没有孩子。
每周五村子里做“parliament meeting”，是一种利用民主投票和讨论来解决纠纷的模拟法庭。在那个村子中心很凉快的大厅下面，有一个人笔录，后面有五六个孩子做“jury”，台下是其他人。大家会讨论一周所有的纠纷。这次说的是在村子里骑自行车的责任问题。一个孩子骑别人的自行车摔伤了。很多人支持骑车，其他人阐述了不同的观点。最后我们决定挂一个意见箱，通过写作来交流。一周后回来，我看到了一群孩子，围在那个袋子面前。当然大大小小的纠纷讨论了好久，不少孩子已经睡着了。
每天和孩子们说晚安，再见。说完晚安之后，他们还是在对面的栅栏后面叫我的名字，我便很安静的让他们去睡觉了。现在他们在一间透风的屋檐下就着很微弱的灯光拼积木。待他们都睡过去，整个森林里就没有了人类的痕迹。这个时候我总是阅读，时常思考，偶尔写作，配着音乐。这周下雨，凉快，但潮湿，蚊子也多。过几个小时我就要回屋睡觉了，风扇调小，现在也从来不用蚊帐了，把门和窗户打开，微弱的灯光下很快就睡过去。除了昨天晚上，我哮喘严重，之后还突然被猫叫给吵醒了。通向二楼的门我都锁了，我很奇怪她是怎么进来的。Ou说她在我进屋之前就一直躲在里面。平时在林里，周末我便去镇子里，一个人在很隐蔽的小旅馆住着。常常呆在屋子里，没有很多虫子，空调和床很舒服，有网络，洗澡水是热的，傍晚很安静，这里有我一切需要的。我会一整天不说话，看电影，阅读。早饭是一趟冒险，可能一无所获。中午饭吃寿司或者711便利。待到夕阳时刻，就踩着一辆很破的自行车去镇子的夜市上觅食。我爱着镇上的夜市，有一条在河边，一条在长途汽车站外面，一条在运动场边。我常常买多，第二天早上正好吃剩饭，经常拉稀。一个人睡去，房子在外面，就只有门锁，有些时候甚至忘记去锁。这时就开始想念 Moo Baan Dek. 我在这里的任何时候都感觉无比安全。之后是尊重，幸福，还有淡淡的快乐，悲伤，不满和焦躁。平静的一切都来之不易，我比想象中更沉醉这种生活。和每个孩子打招呼，说晚安，我们一生能有多少幸福？
父亲，母亲，以及工作者们精神上都很年轻。周末便去镇子里，一个人住着。 燃烧生命
昨晚在我做噩梦的时候我梦到了Moo Baan Dek. 我给予他们每个人拥抱，呼唤着他们的姓名。这些孩子们长大后都不会记得我，因为我只是在他们永恒的童年中一闪而过，但对我来说却是对于残缺的幸福的补全。他们在我低落害怕的时候驱散那些不安，在我沉沦的时候给予我回忆，我很感动和满足。
平静的在温凉的夜晚燃烧，稍微喝一点酒，享受夜晚，长大的权力
坐在橙色的车厢里，之后就是在鲜艳颜色的玻璃海洋里漂流。闷热难耐，偶尔的风吹过。空气混杂着木柴，青草，重视有炎热的风。
世界宜居指南：
知识量统计使用KC，以及结构复杂度来计算。数学以及计算机科学，人工智能最高。文科最高的是法学。
结构复杂度迁移至化学，生物构成。同时比较结构复杂度：历史路径循环，以及稀有程度：数数量，来推断是否有自然/人工选择发生。这个选择，是否也能包括智慧体对于生活中每一个细节的选择？ 选择必定意味着现象的涌现。本质上是因果逻辑推断。但是选择是否意味着意识的产生？也就是说这个理论应用于神经学有意义吗？ 那迁移到语言学呢？
回归纯朴的哲学。Which is the philosophy to-the-earth?
人之间的交通单纯
愿全人幸福，或者至少安稳的度过大部分时光。我们的物质需求少，精神需求多。
但那些大多数人，却没有被那些少数人记住。那些少数人在经历人生的“意义危机”，而那些多数人不应该变成那臆想当中的对比角色。曾不去关心，不去在意，冷漠又温柔。但那只是说明我本身是个无聊冷漠的人。要做出改变。 《路边的藤井》
有人从书中获取知识，我从网站上学习。前者是一个漫长的提炼过程，后者是一个快速，但无法消化的过程。某一个处在两者中间的点才是最优的。
我的好友，孙，提议我记录，整理下来Moo Baan Dek 的这次经历，写成一个比较客观全面的report. 很好的想法。
The first step is to list out the main topics, and then subtopic keywords.
Then bold the keywords that deserve a new story on its own, and start grouping. Try to reduce those to less than 6
reorder the hiearchy
fill-in details
linkage in between stories
all in the meanwhile: forming a coherent, reasonable knowledge structure and worldview
How does a school functions mainly on donations?
Psychology of kids: short term volunteers
the economical structure of thailand
religion
language
race
politics
education v.s. politics and culture
advertisement
泰国的乡村以及城市
贫富差距
贫穷
家庭关系
如何消除自己的无力感？
如何打破人的惰性？
what’s my role, as a volunteer spreading English knowledge, in a thai country?
why is there little volunteers and donations?
How independent, is Moo Baan Dek?
Sep 7 ～ 11：
new things, learning happens every day.
memory-semi conductors: 忆阻器。Using this memrister, you are able to do forward propogations really fast, but can’t do derivation/back propogation? … weird
Qinghua best in this, using it as neurolinks for brains.
I also met a person who solely support his family by investing in stocks. Poor coordination, poor dress, yet sitting comfortably, and has passion when he talks about stocks. Old yellow desktop on a large desk, nothing else in that large living room. Not much furnatures, european styles. And lots of books.
Dealing with digits. Intellectual but meaning diminishing. courageous. I have some idea how my thing might work. Still using traditional ways to trade, late for updates on world tech.
It’s not as good as you think, being a pro trader. Sailing in a sea of unknown future, yet no exploration. Rather should buy ETf or sth.
Sep 15: the biggest day of investment and startup and biotech giants: AI for science, still the very trendy thing.
Chinese economic trend as a whole is not as promising, but the main reason is its heavy industries are declining, exports shrinking, and the best alternative is on its way. The stale money flow incentivise more tech-savvy start ups, not less.
Angel investors should at least consider millions, not a few hundred thousands. The quantity speaks its sincerity. The start up env in china isn’t as good as the silicon valley.
Resource aggregation is the most important quality. aggregate functions, talents, money…
The college graduates in china is barely active enough to push new.
Profit, can wait. But idea, the pipeline, has to be thought out carefully. Time is a really big factor and risk.
Best case scenario when you have both leadership and tech knowledge.
He was
biotech stuff:
company XXXXXX is founded upon the best bio tech of this world, D2P, from SD. crispre came, then transfer this into protein, found that it might work, then tried for almost 10 years, RD. without any products. recently finally done, but still haven’t profited yet.
The problem of predicting/identifying/classifying the DNA sequence that will produce the same protein, but with very different expression speed / volumn, isn’t solved, not even close. There is not enough data. Expensive. Should be algorithmic-dependent, not data dependent.
The predictions using co-variance of two sites, by acknowledging their high-than-usual coordinates or spacing relationships, accross species with proteins of the same function, is widely applied for alphafold predictions
Biotech, or comp bio, could perhaps, be analysed or viewed as the extension of some more general science principles… Categories are the same, morphisms exists accross fields.
data driven, and AI for science.
microscopic things, ultimate reason for difficulties in researching.
high-throughput devised
alphafold didn’t solve the entire omic problem. Only the main chain is predicted, side chain can take many forms, still no conclusions.
Take a look at our midway, around Day 3~4, project structure.
. ├── app/ │ ├── backend/ │ │ ├── api-gateway/ │ │ ├── services/ │ │ │ ├── user-svc/ │ │ │ └── collaboration-service/ │ │ ├── cache/ │ │ ├── db/ │ │ └── logs/ │ └── frontend/ │ ├── src/ │ │ ├── app/ │ │ │ ├── dashboard/ │ │ │ ├── login/ │ │ │ ├── preview/ │ │ │ ├── reader-download/ │ │ │ ├── reader-translating/ │ │ │ ├── reader-workspace/ │ │ │ ├── translate/ │ │ │ ├── translate-editor/ │ │ │ ├── translating/ │ │ │ ├── user-type/ │ │ │ └── workspace/ │ │ ├── components/ │ │ │ └── ui/ │ │ ├── context/ │ │ ├── hooks/ │ │ ├── lib/ │ │ └── styles/ │ ├── public/ │ └── .next/ ├── backend/ │ ├── api/ │ ├── architecture/ │ ├── schemas/ │ └── services/ ├── config/ ├── database/ ├── docs/ ├── Documentations/ ├── frontend/ │ ├── api-gateway/ │ ├── translate-editor/ │ └── user-guides/ ├── infrastructure/ ├── logs/ ├── nextjs-app/ ├── scripts/ ├── services/ │ └── user-service/ ├── Test/ │ └── data/ └── tools/ ├── experiments/ │ └── PDF_to_DOCX/ └── scripts/ └── legacy/
The project appears to be a web application with:
algorithms
software
backtesting