diff --git a/content/posts/2024-09-13_growth.md b/content/posts/2024-09-13_growth.md index db590e3..fa408d2 100644 --- a/content/posts/2024-09-13_growth.md +++ b/content/posts/2024-09-13_growth.md @@ -1,203 +1,206 @@ --- -title: "摘要《經濟成長：歷史與回顧》" +title: "Review: Growth: A Reckoning" date: 2024-09-13T00:09:10+08:00 draft: false images: - images/monkey_sails.png --- +{{< translation type="machine" from="Chinese" />}} + ![](images/monkey_sails.webp) -感謝 Yahsin Huang 和 Anton Cheng 審閱與回饋。 +Thanks to Yahsin Huang and Anton Cheng for review and feedback. + +After reading various books in humanities and social sciences, I've discovered some tendencies. It seems many topics ultimately converge on economic growth. -看了一些人文社科的書之後，我發現了一些傾向。似乎很多主題最後都會收斂到經濟成長。 +- [Four Hundred Years of Taiwan's Economy](https://ketagalanmedia.com/2024/05/20/four-hundred-years-of-taiwans-economy/) discusses Taiwan's economic growth history +- [Poor Economics](https://www.amazon.com/Poor-Economics-Radical-Rethinking-Poverty/dp/1610390938) examines why extremely poor people cannot escape poverty through accumulating capital on their own. The underlying theory is basically an economic growth model, exploring what "traps" appear on the growth path that keep poor people in poverty traps. +- [Power and Progress](https://shapingwork.mit.edu/power-and-progress/) also has an economic growth model at its core, but emphasizes the parts where people can direct technological development. The authors lean left economically. +- I've watched many interviews and read blogs by John Cochrane, mainly to learn his monetary theory. He's an ultra-right economist. He believes economic growth is the holy grail of economics, and he can only contribute some trivial monetary theory. +- [Working to Live](https://www.amazon.com/Work/dp/152660499X). The author, more of an anthropologist, advocates for a degrowth direction. +- I haven't read Capital in the Anthropocene, but it seems to have high discussion levels in Taiwan. The author also advocates degrowth. -- [[台灣經濟四百年]] 在講台灣的經濟成長史 -- [[Poor Economics]] 窮人的經濟學，在講為什麼極度貧窮的人沒辦法透過自行累積資本而脫離貧窮。背後的理論基本上是一個經濟成長模型，並探討成長路徑上出現哪些「陷阱」讓窮人陷入貧窮陷阱。 -- [[Power and Progress]] 權力與進步。背後也是個經濟成長模型，但強調人們能主導科技發展的部分。作者算偏左經濟學家。 -- 看了 John Cochrane 的很多訪談和部落格。主要是想學他的貨幣理論。他算超右經濟學家吧。他認為經濟成長是經濟學聖杯，他只能貢獻一些貨幣理論這種支微末節。 -- [[為工作而活]] 。作者偏人類學家，則是主張棄成長的方向。 -- 我沒看過人類世的資本論，但似乎台灣討論度很高。作者也是主張棄成長。 +The general impression is that economists mostly support economic growth. Often economic growth is treated as the solution to all problems. Sociologists, anthropologists, and environmentalists tend toward degrowth. -大致的印象是如果是經濟學家，大多偏向支持經濟成長。很多時候經濟成長當成是一切問題的解方。社會學家、人類學家、環保人士就會偏棄成長方向。 +## Economic Growth: History and Review -## 經濟成長：歷史與回顧 +Recently I saw the book [Growth: A Reckoning](https://www.danielsusskind.com/growth-a-reckoning) in a bookstore. The author is Daniel Susskind; Taiwan has published his previous book "A World Without Work." I was initially hesitant about whether to really spend time on the topic of economic growth, but bought Growth when I saw it in a second bookstore. I find this book quite efficient—it answered many questions I was curious about, exceeding expectations. -最近在書店看到了 [Growth: A Reckoning](https://www.danielsusskind.com/growth-a-reckoning) 這本書。作者 Daniel Susskind ，台灣有出過他的前一本書「不工作的世界」。本來還在猶豫是否真的要花時間去看經濟成長這個題目，Growth 這本書在第二家書店看到的時候才買下去。我覺得這本書算蠻有效率的，回答了很多我心中好奇的問題，數量超出預期。 +Although the book's title is economic growth, that's more an economist's term. For readers concerned about technology and the future, I think it's an excellent read. Economic growth tells you the priorities among technology, ideas, capital, machinery, labor, education, and so on. The book's writing style breaks out of the circle, assuming readers have little economics knowledge and explaining core concepts with very fluid prose. It also critiques degrowth supporters on the left and full free market/free trade supporters on the right. Finally, it criticizes the longtermism popular in Silicon Valley circles. -這本書雖然書名是經濟成長，但那比較是個經濟學家的詞。對於關心科技與未來的讀者，我覺得是一個很棒的讀物。經濟成長會告訴你：科技、思想、資本、機器、勞力、教育等等這些事情的優先順序為何。這本書的寫法也是個出圈之作，書裡假設讀者有很少的經濟學知識，對於經濟學背景知識都用很流暢的文筆解釋核心概念。並且左打棄成長支持者，右踢全自由市場、自由貿易支持者。最後也對矽谷圈流行的長期主義做了批評。 +As Nobel Prize winner Robert Lucas (who died last year) said: "Once one starts to think about economic growth, it is hard to think about anything else." -就如去年剛過世的諾貝爾經濟學獎得主 Robert Lucas 所說：「一旦人們開始思考經濟成長，就很難再去想其他的事情了。」 +The author believes pursuing economic growth is a very new, mysterious, and dangerous human activity. New because sustained stable economic growth only began in the second half of the 20th century. Mysterious because we don't know why past growth succeeded. Dangerous because economic growth has great potential but also great costs. -作者認為，追逐經濟成長是一件很新、很神秘、也很危險的人類活動。很新因為持續穩定的經濟成長是二十世紀下半才開始。很神秘是因為我們不知道為什麼以前的成長能成功。很危險是因為經濟成長有很大的潛力也有很大的代價。 +For economists, in all of human history, only three things matter. The first 300,000 years were a great stagnation without growth. Then modern economic growth began 200 years ago, though unstable. Only about 100 years ago did we have sustained stable growth. -對經濟學家而言，人類歷史這麼長，只有三件重要的事。前三十萬年是沒有成長的大停滯。然後兩百年前開始有現代的經濟成長，不過不太穩定。接著在大約一百年前才有持續穩定的成長。 +## Under Great Wars, Economic Growth Becomes Popular -## 大戰之下，經濟成長成為流行 +Classical economists had no concept of economic growth. The pursuit of economic growth was somewhat accidental. In 1940, to fight World War II, Keynes had to design a system to measure national output to know how many war resources he could raise for Britain, thus creating GDP's predecessor. In 1933 America experienced the Great Depression, and Kuznets also designed a system to measure national output to understand how bad the Depression really was. -古典經濟學家是沒有經濟成長的概念的。經濟成長的追逐有點是個意外。1940 因為要打世界大戰，凱因斯必須要設計一套衡量國家產出的指標，來知道他能為英國籌多少戰爭資源，因此設計出了 GDP 的前身。 1933 美國那邊發生了大蕭條，Kuznets 也去設計了一套衡量國家產出的指標，來暸解到底大蕭條有多慘。 +Why did no one think to do something similar before 1930? Because designing national wealth indicators was dangerous. Those who told the truth got their heads chopped off. -為什麼在 1930 以前都沒人想做類似的事情？因為設計國家財富指標是一件危險的事。說出真相的人都被砍頭了。 +The main difference between their two designs was that Keynes, raising money for war, included military spending. But Kuznets considered military spending evil and shouldn't be included in the indicator. America chose Keynes's design in 1940—after all, during wartime, defense is very important. -他們兩種設計最主要的差別是，凱因斯是要為戰爭籌錢的，所以有把軍事支出算進去。但 Kuznets 認為軍事支出是個邪惡的支出，不該列在指標裡面。美國在 1940 選擇了凱因斯的設計，畢竟戰爭期間，國防非常重要。 +During the Cold War, the US and USSR had to compete on resource output, comparing GDP daily. Another factor pushing GDP internationally was the Marshall Plan: countries receiving aid had to establish statistical indicators to measure aid effectiveness. -冷戰之間，美蘇必須競爭資源產出，每天都在比較 GDP 。另外一個把 GDP 指標推向國際的是馬歇爾計畫：要領補助的國家必須建立統計指標，以便衡量補助的效果。 +Thus under hot and cold wars, suddenly all countries worldwide had GDP indicators. -因此熱戰冷戰之下，一下子全世界各國都有 GDP 指標了。 +Economic growth indeed brings many benefits. Economic benefits include reducing poverty. Non-economic benefits include healthier lives, more education, and greater happiness. -經濟成長的確會帶來很多好處。經濟上的好處是減少貧窮。非經濟上的利益是更健康的生活、人們受更多教育、人們更快樂。 +For national leaders, focusing on GDP growth has many conveniences. Under economic growth, the unemployed find work, and economic benefits from growth can pay for other social issues. Another benefit is that focusing on growth can delay discussion of all social problems and avoid social conflict. -對於國家的領導人而言，專注於 GDP 的成長有很多方便之處。經濟成長之下，失業的人有工作，成長帶來的經濟利益可以用於支付其他的社會議題。另外的好處就是專注成長可以延遲所有社會問題的討論，迴避社會衝突。 +But economic growth also has many costs, including climate impact, increased inequality, and threatened jobs. -但經濟成長也有許多代價。包含對氣候的衝擊，不平等的增加，工作受到威脅等等。 +## Balancing Economic Growth and Its Costs -## 平衡經濟成長與代價 +So what does the author think should be done to solve or balance the benefits and costs of economic growth? -那作者認為要怎麼解決或平衡經濟成長的好處與代價？ +Before that, let's see what approaches the author disagrees with: -在那之前，先來看作者不同意哪些做法： +Some think since GDP overly focuses on material production without considering climate impacts and inequality, why not design a better indicator and everyone pursue that new indicator? The problem is that indicator design buries moral questions in technical details. For example: How much should we value climate impact? This is a moral question, but in indicator design, it might just become a parameter weight. -有人認為既然 GDP 太著重物質的生產，沒考量到氣候的衝擊和不平等這些問題。那不如設計一個比較好的指標，然後大家追逐那個新指標？這樣的問題是，指標的設計會把一堆道德問題埋到一堆技術細節裡。例如：氣候的衝擊我們應該多重視？這是個道德問題，但在指標設計中，可能只是變成權重的一個參數。 +Therefore the author believes that while GDP's technical problems need time and effort to solve, we shouldn't expect to design a perfect new indicator and just focus on maximizing that number. The author thinks a dashboard approach, listing various issue indicators for people to discuss, is better. -因此作者認為，的確 GDP 的技術問題需要花時間和力氣解決。但我們不應該期待能設計出一個完美的新指標，然後專注把數字衝高就好。作者認為一個儀表板的做法，把各種議題的指標列出來讓人們去討論的做法比較好。 +The second approach is degrowth. The author spends effort laying out the historical context of degrowth thinking. But the most popular myth is "Earth's resources are limited, how can they support infinite economic growth?" The problem with this myth is viewing economic activity with outdated perspectives, only seeing the material part. -第二種做法就是棄成長。作者花了些力氣去鋪陳棄成長思維的歷史脈絡。不過人們最流行的迷思就是「地球資源有限，怎麼能支援無限的經濟成長？」這個迷思的問題是用老舊的觀點在看經濟活動，只看到物質的部分。 +Economic growth isn't driven by using more limited resources, but by discovering more ways to better use limited resources. Paul Romer has a vivid description: ingredients in a kitchen are limited, but recipes these ingredients can form are nearly infinite. 300 ingredients can combine into more recipes than atoms in the universe, even without considering different proportions. People might say most recipes would be useless, but the explorable space remains huge. The curse of dimensionality is a blessing here. -經濟成長並不是由運用更多的有限資源去驅動，而是藉由發現更多善用有限資源的方式去驅動。 Paul Romer 有個生動的描述：廚房裡的食材是有限的，但這些食材能組成的食譜近乎無限。300 種食材能組合出的食譜比宇宙中的原子還要多，這還是在不考慮不同食材組成比例之下。人們也許會說這裡面大多數的食譜都是沒作用的，但能探索的空間仍然很大。高維度的詛咒在這邊是種福氣。 +Abandoning the full pursuit of growth has many levels—you could choose to slow growth, but degrowthers often choose the most extreme negative growth route. -放棄全力追逐成長也有許多層次，你可以選擇減緩成長，但棄成長者往往選擇最極端的負成長路線。 +Therefore, the author's biggest criticism of degrowthers is lack of imagination, choosing the most extreme solutions and utopian project routes. But the author still thinks degrowthers' arguments deserve engagement, partly because their arguments are popular. Also, the current situation somewhat pursues economic growth at any cost, while degrowthers abandon growth at any cost. We can choose a middle path between these. -因此，作者批評棄成長主義者最大的問題就是缺乏想像力，選擇最極端的解決方案和烏托邦專案路線。但作者仍然認為棄成長主義者的言論值得面對，一方面當然是因為他們的言論很流行。另一方面，現在的情況有點是不計代價追逐經濟成長，而棄成長主義者則是不計代價棄成長。我們可以在這之間選個中庸之道。 +## Liberating the Production of Knowledge and Ideas -## 解放知識與想法的生產 +Having discussed new indicators and degrowth, we can discuss what the author considers the right path for growth. -討論完設計新指標和棄成長，我們可以來談作者認為成長的正途是什麼了。 +But first we need some background on what efforts humanity has made and what mistakes we've committed in understanding economic growth. The author summarizes four basic approaches: -但在那之前又要補一下一些背景，到底人類在理解經濟成長的過程中，做了哪些努力又犯了哪些錯。作者歸納人們基本上有四種作法： +The first is writing ten-thousand-word essays trying to organize a narrative like social theory—this approach is a major failure, with no successful predictions. -第一種是學社會理論寫萬字長文試圖整理出一個敘事，這種作法是大失敗，沒有預測是成功的。 +The second and most successful is concise mathematical models. Harrod and Domar discussed combinations of labor and capital, physical capital investment and economic growth. Solow and Swan improved the former model, adding technological factors. This is important—humans initially trapped in great stagnation because labor productivity has diminishing marginal returns. One more person means two more hands to work, but also one more mouth to feed. The extra hands produce diminishingly, unable to feed the extra mouth, trapped in the Malthusian trap. Therefore more people doesn't help—the game of economic growth requires increasing the average economic fruits per person to count as growth. -第二種最成功的是簡潔的數學模型。Harrod 與 Domar 去討論勞力與資本的組合，物理資本投資與經濟成長。Solow 與 Swan 改良前者模型，加入了技術要素。這很重要，人類一開始陷在大停滯裡，是因為勞力生產力會邊際效應遞減。多生了一個人會多雙手幫忙工作，但也多張口要餵。多出來的手生產遞減，餵不了多出來的口，就會陷在馬爾薩斯陷阱裡。因此人多沒有用，經濟成長的遊戲規則是必須要讓每個人平均分到的經濟果實變多了，才算有成長。 +Technology's power is that it changes production methods, offsetting diminishing marginal productivity of labor and capital. -科技厲害的地方就是他改變了生產方式，抵消了勞力與資本的邊際生產力遞減。 +Robert Lucas and Paul Romer further discovered the importance of human capital and skills. Investing in education also improves production efficiency. -Robert Lucas, Paul Romer 又進一步發現人力資本與技能的重要性。投資在教育上也能改善生產效率。 +The third is data analysis, but the author says every data analysis result has multiple interpretations. This approach doesn't contribute much. -第三種是資料分析的做法，但作者說每種資料分析的結果都有多種詮釋。這個做法貢獻也不大。 +The fourth pursues fundamental approaches. Are there fundamental reasons causing economic growth? Jared Diamond's geography, Max Weber's work ethic, Douglass North's institutions, Daron Acemoglu and James Robinson's extractive and inclusive institutions belong to this category. The author most likes the combination of Joel Mokyr's "A Culture of Growth" and Paul Romer's intangible knowledge model. Though the author notes Mokyr and Romer don't acknowledge each other, never citing each other's research. -第四種是追求基本面的做法。到底有沒有根本面的原因造成經濟成長？賈德戴蒙的地理學、馬克斯韋伯的工作倫理、Douglass North 的制度，Daron Acemoglu 和 James Robinson 的榨取性和廣納性制度等等屬於這類。其中作者最喜歡 Joel Mokyr 的 A culture of Growth 與 Paul Romer 無形知識模型的組合。不過作者也說 Mokyr 與 Romer 王不見王，互相都不引用對方的研究。 +With this background, the author believes pursuing economic growth means asking how to pursue technological development and increase production of knowledge and ideas. The author thinks four things need addressing: -有了這些背景之後，作者認為追逐經濟成長，就是在問怎麼追逐科技發展，和怎麼增加知識與想法的生產。作者認為有四件事情要處理： +First is asking who owns and controls knowledge and ideas: intellectual property rights. IP protection provides inventors incentives to invent. Too short protection and no one wants to invent; too long and knowledge can't circulate to have impact. The author notes current IP systems have problems of being outdated (protection too long), excessive (protection overly granted), and weaponized (used for anti-competitive purposes). -第一是去問誰擁有和控制知識與想法：也就是智慧財產權。智財保護的用意在於提供發明者發明想法的誘因。保護得太短則沒人想發明，保護得太長知識又無法流通發揮影響。作者指出目前的智財制度有老舊（保護太長）、氾濫（保護濫發）、和武器化（用來做反競爭用途）的問題。 +Second is increasing R&D investment. Currently worldwide average R&D investment is 1-4% of GDP, with the top three: Israel about 6%, Korea 5%, Taiwan 4%. Countries and large companies are somewhat unfair, but big tech companies' R&D as percentage of profit: Alphabet 15%, MSFT 13%, FB 21%. The author thinks countries should raise R&D percentages even higher. -第二是要增加研發的投資。目前全世界平均對研發的投資是 GDP 的 1%~4% ，其中前三名：以色列大概 6%，韓國 5% ，台灣 4% 。國家和大企業比較不公平，但大科技公司的研發佔獲利比： Alphabet 15%, MSFT 13%, FB 21% 。作者是認為國家應該要再把研發的比例拉更高。 +Third is getting more people involved in R&D. Research shows low-income people are less likely to become inventors when they grow up. If this problem can be alleviated, it can both solve opportunity inequality caused by income and promote growth through more people participating in R&D. -第三是要讓更多人們參與研發。研究顯示低收入的人們在長大比較不容易成為發明家。如果這個問題可以緩解，一方面可以解決因收入造成的機會不平等，另一方面更多人參與研發可以促進成長。 +Fourth is increasing researcher efficiency. Our research efficiency has decreased, the lowest-hanging fruit seems picked: -第四是要增加研究員的效率。我們研究的效率降低了，最低的果實似乎摘完了： +- Moore's Law has indeed held steady from 1970 to today, but achieving the same effect today requires 18 times more human effort than 1970. +- Discovering the God particle used 5,000 researchers, with a 33-page paper having only nine pages of actual content, the remaining pages listing author names. +- University administrative bloat: Harvard has 7,000 undergraduates but 7,000 administrators. +- Researchers spend 40% of their time applying for funding. +- Recent surveys show 80% of researchers, constrained by funding conditions, research topics they don't like. -- 摩爾定律的確從 1970 至今穩定發揮，但要達成同樣的效果，今天要比 1970 花 18 倍的人力。 -- 發現上帝粒子用了 5000 個研究員，33 頁的論文只有九頁是真正內容，剩下的頁面都在列出作者名稱。 -- 大學的行政人員膨脹：哈佛有 7000 大學部學生，但有 7000 行政人員。 -- 研究員花 40% 時間在申請經費。 -- 近期調查顯示 80% 研究員在受申請經費條件的限制之下，做不喜歡的研究主題。 +The author mentions that improving research efficiency might not be possible through humans alone. For protein folding, previously researching one protein's 3D shape used up almost an entire PhD career, and today only 17% of known proteins' 3D shapes are known. But in 2022 DeepMind's AI discovered 100 million protein 3D shapes in one go. -作者提到，要改善研究效率，可能沒辦法純靠人了。以蛋白質折疊問題而言，以往研究一個蛋白質的立體形狀幾乎就是用掉一個 PhD 的研究生涯，在今天也只有 17% 已知的蛋白質的立體形狀。但 2022 DeepMind 的 AI 一口氣就發現一億種蛋白質立體形狀了。 +But the most interesting part of the book is what the author thinks cannot boost economic growth much: -不過書裡最有趣的是，作者認為哪些作為是無法提升經濟成長太多的。 +- Building bridges and roads, more infrastructure: Large projects waste lots of money. And this thinking falls into capital-oriented thinking. In growth models, marginal utility of capital increases diminishes. So theoretically unfeasible. +- Better land use and urban planning: Paul Romer likes cities because he thinks they let many people meet in the same place, promoting knowledge exchange, and knowledge is the nutrient for economic growth. But the author thinks there's no empirical data support—with remote work technology prevalent today, is physical meeting really that important? +- Increasing education: Few countries can get 90% of people through middle school, 50% through college. The author thinks the past "century of human capital" can't happen again. I think this is the author's most interesting view. I also thought educational improvement should have lots of potential. But even if we could double people's abilities? Or even ten times? Can it compare to AI productivity improvements like protein 3D structures? -- 造橋鋪路，蓋更多基礎建設：大型建設造成很多金錢浪費。而且這個思路陷入資本導向的思維。在成長模型中，資本增加的邊際效用會遞減。所以理論上不可行。 -- 更好的土地利用與都市規劃：Paul Romer 喜歡都市，因為他認為這是大量的人能在同一個地方見面，促進知識交流，而知識就是經濟成長的養分。但作者認為沒有任何實證資料佐證，在現在遠端工作科技盛行之下，實體見面真的那麼重要嗎？ -- 增加教育：很少國家能讓 90% 以上的人上完初中，50% 的人上完大學。作者認為過去「人力資本的世紀」已經無法再發生了。我認為這是作者最有趣的觀點。我本來也覺得教育的改良應該有很多潛力。但即使我們有能力把人的能力再提升兩倍？或甚至十倍。能比得上蛋白質立體形狀那種 AI 的生產力進步嗎？ +## The Direction of Growth -## 成長的方向 +The author notes people easily think of economic growth like a locomotive. Growth supporters think we should accelerate, degrowthers think we should slow down or reverse. But this metaphor's problem is viewing growth like railroad tracks with only one direction, only forward or backward. -作者指出，人們很容易把經濟成長想成一台火車頭。成長的人認為要加速，棄成長的認為要減速或倒車。但這個比喻的問題在於把成長看成像鐵軌一樣只有一個方向，只能前進或是後退。 +A better metaphor is a ship in the ocean—the ship can advance in different directions, and when looking at growth we also need to choose which direction to advance. -比較好的比喻是像汪洋中的一條船，船有不同的方向可以前進，而我們在看成長時也要選擇往哪個方向前進。 +An important observation: the same 5% growth in Britain—today's 5% produces very different things than 5% a hundred years ago. -一個重要的觀察是：同樣在英國 5% 的成長，現在的 5% 與和在一百年前的 5% 生產出來的東西是非常不一樣的。 +The difference of course is we now have completely different technology. But is technological development something we can control? -這當中的差別當然是我們現在有完全不一樣的科技。但科技的發展難道是我們有辦法控制的嗎？ +The author distinguishes two concepts: induced technological development and directed technological development. -作者區分了兩種概念：誘發性的科技發展和引導性的科技發展。 +An example of induced technological development is Britain's Industrial Revolution. Why didn't people in nearby Germany, France, and Belgium develop the Industrial Revolution at the same time? Because Britain at that time had both high wages and low fuel costs, making it profitable to invest in improving steam engines to replace workers. Similar examples include: Japan's advanced care robots because of large elderly populations and low immigration. And China's industrial robots, because the low-wage environment manufacturing previously relied on no longer exists. -誘發性的科技發展的例子像是英國工業革命。為什麼同樣時期，鄰近的德國法國和比利時地區的人沒發展出工業革命？那是因為當時的英國同時有高工資和低燃料成本的環境，使得投入蒸汽機的改良去取代勞工有利可圖。類似的例子還有：在日本照護機器人很先進，因為老人人口多，移民又很低。以及中國的工業機器人，因為製造業以往倚賴的低工資環境已經不在。 +People's understanding of directed technological development has greatly increased recently, mainly through [[Daron Acemoglu]]'s research. The main arguments are roughly the content of the Power and Progress book. Generally, those factors or incentives inducing technological development aren't completely beyond our control. Through taxes and subsidies, regulations, and changing social norms, we can direct technological development's direction. -人們近年對引導性的科技發展的理解大幅增加。主要是 [[Daron Acemoglu]] 的研究。主要論點大概是權力與進步這本書的內容。大致而言，那些誘發科技發展的因素或誘因，並非我們完全沒有能力控制的。透過稅與補助、法規、與社會規範的改變，我們有能力引導科技發展的方向。 -- 疫情期間的隔離政策可以視為，公司要聘人進辦公室的工資成本無限大，因此誘發了 Zoom skype 等遠端工作的科技與文化發生。因為這個政策的因素，催生的這些科技的進程。 -- 疫情初期，疫苗本來預期要十年才能開發出來。但我們八個月之後就有 AZ 和 BNT 。幾個月之後就有莫德納和嬌生。這也是工資成本無限大促發疫苗開發的強烈誘因。 -- 壞的方面是，很多公司因應疫情，把那些需要見面容易傳染的工作自動化了。這造成很多勞工的困境。 +- Pandemic isolation policies can be viewed as making the wage cost of hiring people into offices infinitely high, thus inducing technologies and cultures like Zoom and Skype for remote work. These policies catalyzed these technologies' progress. +- Early in the pandemic, vaccines were expected to take ten years to develop. But we had AZ and BNT after eight months. Moderna and J&J came months later. This was also infinite wage costs creating strong incentives for vaccine development. +- On the negative side, many companies automated jobs requiring face-to-face contact prone to infection in response to the pandemic. This created difficulties for many workers. -引導性科技發展的理論其實受到許多左派右派的挑戰。 +The theory of directed technological development faces challenges from both left and right. -右派如市場基本教義派的如胡佛研究所就會說，引導性科技發展是一種政策干預市場價格機制正常運作的方式。作者的反駁是，對我們作為「消費者」身份時，市場機制會反映我們在乎的事情。但我們作為「公民」身份時，市場機制難以反映我們的顧慮。作者也回應引導性科技發展並非中央規劃。 +The right, like market fundamentalists at the Hoover Institution, say directed technological development is policy intervention disrupting normal market price mechanisms. The author's rebuttal is that when we act as "consumers," market mechanisms reflect what we care about. But when we act as "citizens," market mechanisms struggle to reflect our concerns. The author also responds that directed technological development isn't central planning. -左派則批評引導性科技發展仍然是利用財產權、誘因、市場機制去鞏固經濟成長和資本主義，像是請鬼拿藥單。作者則是提醒他們忘記左派文獻的根源：經濟成長並不是資本主義獨有的東西。在以前社會主義社會，他們重視成長更甚於資本主義社會，並以物質產出的數量為傲。 +The left criticizes directed technological development as still using property rights, incentives, and market mechanisms to consolidate economic growth and capitalism, like asking ghosts for prescriptions. The author reminds them they forget leftist literature's roots: economic growth isn't unique to capitalism. In former socialist societies, they valued growth even more than capitalist societies and took pride in material output quantities. -作者引用 Derek Parfit ：我們目前處於一個歷史的轉捩點。在過去兩百年間科技高速變化。我們會有更大的能力去改變環境、甚至是自己或後代。如果我們在接下來幾百年做明智的行動，人類可以度過這個最危險最關鍵的時期。 +The author quotes Derek Parfit: We are at a turning point in history. Technology has changed rapidly over the past 200 years. We will have greater ability to change the environment, even ourselves or our descendants. If we take wise actions in the coming centuries, humanity can survive this most dangerous and critical period. -作者說，如果我們有更多能力去改變未來，那代表過去能給我們的經驗就變少了。例如：以前的人批評「綠色成長（Green Growth）」不可能發生，但以現在的資訊來看，這在未來極有可能。 +The author says if we have more ability to change the future, that means the past can give us less experience. For example: people previously criticized "Green Growth" as impossible, but with current information, this is extremely likely in the future. -如果科技發展真的是我們有辦法引導其方向，這也代表我們對我們選擇的方向有道德責任。 +If technological development really can be directed, this also means we have moral responsibility for our chosen directions. -## 成長的取捨與道德責任 +## Growth's Trade-offs and Moral Responsibility -前面提到，經濟成長有氣候的衝擊，不平等的增加，工作受到威脅等等代價。我們要如何在眾多重視的價值中做取捨？ +As mentioned, economic growth has costs like climate impact, increased inequality, and threatened jobs. How do we make trade-offs among many valued priorities? -首先作者先討論，取捨是必要的嗎？也許我們可以先想，有沒有避免抵換的方式？（Pareto 改善）再者，如果完全避免是不可能的，那有沒有弱化抵換的方式？窮盡這些選項之後，我們可以再來談取捨。 +First the author discusses: are trade-offs necessary? Maybe we can first think if there are ways to avoid trade-offs (Pareto improvements)? If complete avoidance is impossible, are there ways to weaken trade-offs? After exhausting these options, we can discuss trade-offs. -- 避免抵換 - - 前面提到如果我們給低收入的人更多機會，這能夠同時改善不平等和增進成長。 -- 弱化抵換 - - 作者花了很多力氣談綠色成長：也就是在不增加環境負擔之下增加經濟成長。 - - 2007 Stern review: 要把 CO2 降到現在的 80% 以下，要花費每年 1% GDP 。 - - 但到了2020 Climate Change Committee。完全消滅排碳的成本只需要 0.5% GDP 。 - - 1976 太陽能價格是每瓦 100 美元。到 2019 變成每瓦 0.5 美元。 - - IEA: 對 2030 年全球使用太陽能的預測。 2021 比 2006 年的預測多 30 倍。預期在 2027 年太陽能超越煤炭和天然氣。 - - 綠色脫鉤：許多國家正在發生 GDP 成長又同時減少排碳的情況。 - - 綠色脫鉤可以說是發明家、企業家等集體努力之下而成。透過文化、制度、稅與補助、法規、及社會敘事與規範共同塑造。 - - 烏俄戰爭在俄斷天然氣之下，間接造成歐洲綠能轉型。 - - 自動化 - - 比爾蓋茲觀察到：為什麼工人要繳稅，但工廠把工人換成機器人之後，機器人不用繳稅。這裡反映的是，勞力和資本同樣是生產要素，但勞力太容易被課稅。 - - 讀者可能會很想對資本課稅，來減少成長對不平等的抵換。但對資本課稅要很小心，一不小心容易得到反效果。 +- Avoiding trade-offs + - As mentioned, giving low-income people more opportunities can simultaneously improve inequality and promote growth. +- Weakening trade-offs + - The author spends much effort discussing green growth: increasing economic growth without increasing environmental burden. + - 2007 Stern Review: Reducing CO2 to below 80% of current levels would cost 1% GDP annually. + - But by 2020 Climate Change Committee: Complete elimination of carbon emissions only needs 0.5% GDP cost. + - 1976 solar price was $100 per watt. By 2019 it became $0.5 per watt. + - IEA: Predictions for 2030 global solar use. 2021 predictions are 30 times higher than 2006 predictions. Solar expected to surpass coal and natural gas by 2027. + - Green decoupling: Many countries experiencing GDP growth while simultaneously reducing carbon emissions. + - Green decoupling results from collective efforts of inventors, entrepreneurs, etc. Shaped together through culture, institutions, taxes and subsidies, regulations, and social narratives and norms. + - Russia-Ukraine war indirectly caused Europe's green transition after Russia cut natural gas. + - Automation + - Bill Gates observed: Why do workers pay taxes, but when factories replace workers with robots, robots don't pay taxes? This reflects that while labor and capital are both production factors, labor is too easily taxed. + - Readers might want to tax capital to reduce growth's trade-off with inequality. But taxing capital requires care—carelessly done easily produces opposite effects. -談完避免和弱化抵換之後，便能來談接受抵換的情況。 +After discussing avoiding and weakening trade-offs, we can discuss accepting trade-offs. -要談成長與環境、不平等等抵換，其實是在討論平衡現在人與未來人的利益的問題。現在成長不足，是留給未來比較少的手段解決問題，但也不能留給未來人一些不平等和污染的爛攤子。 +Discussing trade-offs between growth and environment, inequality, etc., is really discussing balancing present and future people's interests. Insufficient growth now leaves the future fewer means to solve problems, but we also can't leave future people messes of inequality and pollution. -經濟學裡面有個「代際平等（intergeneration equity）」的題目。通常是用折現率這樣一個參數，來表達我們怎麼平衡現在人與未來人的利益。 +Economics has a topic of "intergenerational equity." Usually using a parameter like discount rate to express how we balance present and future people's interests. -作者認為，現在的人與文化的確是短視近利。企業極大化短期報酬來取悅投資人，政府追逐短期成就來取悅選民，而民眾則是被每天的煩惱所困擾。這算是一種極重視短期的極端。 +The author believes current people and culture are indeed short-sighted. Companies maximize short-term returns to please investors, governments pursue short-term achievements to please voters, and citizens are troubled by daily worries. This is an extreme emphasis on the short term. -作為另一個極端的比較，作者提到矽谷最流行的長期主義。（他說是 exteremely helpful mistake）。 +As a comparison at the other extreme, the author mentions Silicon Valley's popular longtermism (he calls it an "extremely helpful mistake"). -長期主義關注幾百萬年、幾千萬年、或幾億年以後。人類已經突破 80 兆人口，在那個最遠的數字，地球已經被燃燒殆盡的太陽給吞噬，人們已經在外太空飄了。 +Longtermism focuses on millions, tens of millions, or billions of years later. Humanity has broken through 80 trillion population, at that furthest number, Earth has been swallowed by the burning sun, people are floating in outer space. -長期主義者認為，我們應該要為那些遠在未來的人謀福利，因為他們人數比我們大太多了。 +Longtermists believe we should work for the welfare of those far-future people, because they vastly outnumber us. -但作者說長期主義者想保護的未來太遙遠了，而且是我們無法想像。我們有可能近期就長壽運動成功，突破長壽逃逸速度（ longevity escape velocity），也有可能上傳我們的心智達到數位永生。但這些事情都已經超越我們的想像，更何況那些幾千年幾億年的東西。 +But the author says the future longtermists want to protect is too distant and beyond our imagination. We might soon succeed in longevity movements, breaking through longevity escape velocity, or upload our minds for digital immortality. But these things already exceed our imagination, let alone things thousands or billions of years away. -作者引用 Regina Rini ：「我們試圖預測數百萬年後星際旅行者的需求，就如同舊石器時代的酋長為二十一世紀的矛頭準備尖銳燧石一樣不切實際。」 +The author quotes Regina Rini: "Our attempts to predict the needs of interstellar travelers millions of years hence are as impractical as a Paleolithic chieftain preparing sharp flints for 21st-century spearheads." -作者還偷酸了一下長期主義的重要支持者 SBF：「結果人類最致命的威脅不是未來的核戰或超級人工智慧，而是在我們時代的一個三十歲捲捲頭髮的億萬富翁」。其經營的 FTX 交易所糟糕的內控造成災難一場。 +The author also takes a jab at important longtermist supporter SBF: "Turns out humanity's deadliest threat isn't future nuclear war or superintelligent AI, but a thirty-year-old curly-haired billionaire in our time." His FTX exchange's terrible internal controls caused disaster. -因為未來科技發展的難以想像，作者認為我們應該擁抱「想像力的謙遜」及「務實的謙遜」，去看待未來對現在的需求。 +Because future technological development is unimaginable, the author believes we should embrace "imaginative humility" and "practical humility" in viewing the future's demands on the present. -所以我們欠未來人什麼呢？作者認為我們應該要比現代人再更加關心未來，但比長期主義者關注再更加近的事情。不能和經濟學家一樣爭論折現率的幾個小數點，但必須爭論各種議題的取捨方向。 +So what do we owe future people? The author believes we should care more about the future than modern people do, but focus on nearer things than longtermists. We can't argue about decimal points in discount rates like economists, but must argue about trade-off directions for various issues. -對於道德問題，作者最後還是丟了「要讓公民決定」這種答案。作者提了一些類似國民法官這種 Mini-publics 的機制，在國家中抽樣人民去為特定的議題找答案。 +For moral questions, the author ultimately throws out the answer "let citizens decide." The author mentions mechanisms like Mini-publics similar to citizen judges, sampling citizens nationally to find answers for specific issues. -## 結論 +## Conclusion -作者基本上談論經濟成長的起源，人們做了哪些研究，並討論經濟成長的好處與代價。在平衡好處與代價之間，並非如現在一昧追逐成長和完全放棄成長這麼極端，而是有許多方向與中間地帶可以取捨。 +The author basically discusses economic growth's origins, what research people have done, and discusses growth's benefits and costs. In balancing benefits and costs, it's not as extreme as blindly pursuing growth or completely abandoning growth, but there are many directions and middle grounds for trade-offs. -我覺得這本書是我最近看了幾樣東西的交匯點。 +I think this book is the confluence of several things I've recently read: -- 我可以理解為什麼 Glen Weyl 要那麼強調報酬遞增的研究（現在諾貝爾獎頒太多給報酬遞減的研究）。 -- 知識、想法在經濟上的重要性。 -- 權力與進步講的引導性科技發展為什麼重要。 -- 我們看到各種新科技要怎麼放到經濟成長的角度去理解。 +- I can understand why Glen Weyl emphasizes research on increasing returns so much (too many Nobel Prizes now go to research on decreasing returns). +- The economic importance of knowledge and ideas. +- Why directed technological development discussed in Power and Progress is important. +- How to understand various new technologies from an economic growth perspective. -另外也看到很多之前沒關注到，重要的事。例如綠能的發展，蛋白質的摺疊。 +I also saw many previously unnoticed important things. For example, green energy development, protein folding. \ No newline at end of file diff --git a/content/posts/2024-09-13_growth.zh-TW.md b/content/posts/2024-09-13_growth.zh-TW.md new file mode 100644 index 0000000..db590e3 --- /dev/null +++ b/content/posts/2024-09-13_growth.zh-TW.md @@ -0,0 +1,203 @@ +--- +title: "摘要《經濟成長：歷史與回顧》" +date: 2024-09-13T00:09:10+08:00 +draft: false +images: +- images/monkey_sails.png +--- + +![](images/monkey_sails.webp) + +感謝 Yahsin Huang 和 Anton Cheng 審閱與回饋。 + +看了一些人文社科的書之後，我發現了一些傾向。似乎很多主題最後都會收斂到經濟成長。 + +- [[台灣經濟四百年]] 在講台灣的經濟成長史 +- [[Poor Economics]] 窮人的經濟學，在講為什麼極度貧窮的人沒辦法透過自行累積資本而脫離貧窮。背後的理論基本上是一個經濟成長模型，並探討成長路徑上出現哪些「陷阱」讓窮人陷入貧窮陷阱。 +- [[Power and Progress]] 權力與進步。背後也是個經濟成長模型，但強調人們能主導科技發展的部分。作者算偏左經濟學家。 +- 看了 John Cochrane 的很多訪談和部落格。主要是想學他的貨幣理論。他算超右經濟學家吧。他認為經濟成長是經濟學聖杯，他只能貢獻一些貨幣理論這種支微末節。 +- [[為工作而活]] 。作者偏人類學家，則是主張棄成長的方向。 +- 我沒看過人類世的資本論，但似乎台灣討論度很高。作者也是主張棄成長。 + +大致的印象是如果是經濟學家，大多偏向支持經濟成長。很多時候經濟成長當成是一切問題的解方。社會學家、人類學家、環保人士就會偏棄成長方向。 + +## 經濟成長：歷史與回顧 + +最近在書店看到了 [Growth: A Reckoning](https://www.danielsusskind.com/growth-a-reckoning) 這本書。作者 Daniel Susskind ，台灣有出過他的前一本書「不工作的世界」。本來還在猶豫是否真的要花時間去看經濟成長這個題目，Growth 這本書在第二家書店看到的時候才買下去。我覺得這本書算蠻有效率的，回答了很多我心中好奇的問題，數量超出預期。 + +這本書雖然書名是經濟成長，但那比較是個經濟學家的詞。對於關心科技與未來的讀者，我覺得是一個很棒的讀物。經濟成長會告訴你：科技、思想、資本、機器、勞力、教育等等這些事情的優先順序為何。這本書的寫法也是個出圈之作，書裡假設讀者有很少的經濟學知識，對於經濟學背景知識都用很流暢的文筆解釋核心概念。並且左打棄成長支持者，右踢全自由市場、自由貿易支持者。最後也對矽谷圈流行的長期主義做了批評。 + +就如去年剛過世的諾貝爾經濟學獎得主 Robert Lucas 所說：「一旦人們開始思考經濟成長，就很難再去想其他的事情了。」 + +作者認為，追逐經濟成長是一件很新、很神秘、也很危險的人類活動。很新因為持續穩定的經濟成長是二十世紀下半才開始。很神秘是因為我們不知道為什麼以前的成長能成功。很危險是因為經濟成長有很大的潛力也有很大的代價。 + +對經濟學家而言，人類歷史這麼長，只有三件重要的事。前三十萬年是沒有成長的大停滯。然後兩百年前開始有現代的經濟成長，不過不太穩定。接著在大約一百年前才有持續穩定的成長。 + +## 大戰之下，經濟成長成為流行 + +古典經濟學家是沒有經濟成長的概念的。經濟成長的追逐有點是個意外。1940 因為要打世界大戰，凱因斯必須要設計一套衡量國家產出的指標，來知道他能為英國籌多少戰爭資源，因此設計出了 GDP 的前身。 1933 美國那邊發生了大蕭條，Kuznets 也去設計了一套衡量國家產出的指標，來暸解到底大蕭條有多慘。 + +為什麼在 1930 以前都沒人想做類似的事情？因為設計國家財富指標是一件危險的事。說出真相的人都被砍頭了。 + +他們兩種設計最主要的差別是，凱因斯是要為戰爭籌錢的，所以有把軍事支出算進去。但 Kuznets 認為軍事支出是個邪惡的支出，不該列在指標裡面。美國在 1940 選擇了凱因斯的設計，畢竟戰爭期間，國防非常重要。 + +冷戰之間，美蘇必須競爭資源產出，每天都在比較 GDP 。另外一個把 GDP 指標推向國際的是馬歇爾計畫：要領補助的國家必須建立統計指標，以便衡量補助的效果。 + +因此熱戰冷戰之下，一下子全世界各國都有 GDP 指標了。 + +經濟成長的確會帶來很多好處。經濟上的好處是減少貧窮。非經濟上的利益是更健康的生活、人們受更多教育、人們更快樂。 + +對於國家的領導人而言，專注於 GDP 的成長有很多方便之處。經濟成長之下，失業的人有工作，成長帶來的經濟利益可以用於支付其他的社會議題。另外的好處就是專注成長可以延遲所有社會問題的討論，迴避社會衝突。 + +但經濟成長也有許多代價。包含對氣候的衝擊，不平等的增加，工作受到威脅等等。 + +## 平衡經濟成長與代價 + +那作者認為要怎麼解決或平衡經濟成長的好處與代價？ + +在那之前，先來看作者不同意哪些做法： + +有人認為既然 GDP 太著重物質的生產，沒考量到氣候的衝擊和不平等這些問題。那不如設計一個比較好的指標，然後大家追逐那個新指標？這樣的問題是，指標的設計會把一堆道德問題埋到一堆技術細節裡。例如：氣候的衝擊我們應該多重視？這是個道德問題，但在指標設計中，可能只是變成權重的一個參數。 + +因此作者認為，的確 GDP 的技術問題需要花時間和力氣解決。但我們不應該期待能設計出一個完美的新指標，然後專注把數字衝高就好。作者認為一個儀表板的做法，把各種議題的指標列出來讓人們去討論的做法比較好。 + +第二種做法就是棄成長。作者花了些力氣去鋪陳棄成長思維的歷史脈絡。不過人們最流行的迷思就是「地球資源有限，怎麼能支援無限的經濟成長？」這個迷思的問題是用老舊的觀點在看經濟活動，只看到物質的部分。 + +經濟成長並不是由運用更多的有限資源去驅動，而是藉由發現更多善用有限資源的方式去驅動。 Paul Romer 有個生動的描述：廚房裡的食材是有限的，但這些食材能組成的食譜近乎無限。300 種食材能組合出的食譜比宇宙中的原子還要多，這還是在不考慮不同食材組成比例之下。人們也許會說這裡面大多數的食譜都是沒作用的，但能探索的空間仍然很大。高維度的詛咒在這邊是種福氣。 + +放棄全力追逐成長也有許多層次，你可以選擇減緩成長，但棄成長者往往選擇最極端的負成長路線。 + +因此，作者批評棄成長主義者最大的問題就是缺乏想像力，選擇最極端的解決方案和烏托邦專案路線。但作者仍然認為棄成長主義者的言論值得面對，一方面當然是因為他們的言論很流行。另一方面，現在的情況有點是不計代價追逐經濟成長，而棄成長主義者則是不計代價棄成長。我們可以在這之間選個中庸之道。 + +## 解放知識與想法的生產 + +討論完設計新指標和棄成長，我們可以來談作者認為成長的正途是什麼了。 + +但在那之前又要補一下一些背景，到底人類在理解經濟成長的過程中，做了哪些努力又犯了哪些錯。作者歸納人們基本上有四種作法： + +第一種是學社會理論寫萬字長文試圖整理出一個敘事，這種作法是大失敗，沒有預測是成功的。 + +第二種最成功的是簡潔的數學模型。Harrod 與 Domar 去討論勞力與資本的組合，物理資本投資與經濟成長。Solow 與 Swan 改良前者模型，加入了技術要素。這很重要，人類一開始陷在大停滯裡，是因為勞力生產力會邊際效應遞減。多生了一個人會多雙手幫忙工作，但也多張口要餵。多出來的手生產遞減，餵不了多出來的口，就會陷在馬爾薩斯陷阱裡。因此人多沒有用，經濟成長的遊戲規則是必須要讓每個人平均分到的經濟果實變多了，才算有成長。 + +科技厲害的地方就是他改變了生產方式，抵消了勞力與資本的邊際生產力遞減。 + +Robert Lucas, Paul Romer 又進一步發現人力資本與技能的重要性。投資在教育上也能改善生產效率。 + +第三種是資料分析的做法，但作者說每種資料分析的結果都有多種詮釋。這個做法貢獻也不大。 + +第四種是追求基本面的做法。到底有沒有根本面的原因造成經濟成長？賈德戴蒙的地理學、馬克斯韋伯的工作倫理、Douglass North 的制度，Daron Acemoglu 和 James Robinson 的榨取性和廣納性制度等等屬於這類。其中作者最喜歡 Joel Mokyr 的 A culture of Growth 與 Paul Romer 無形知識模型的組合。不過作者也說 Mokyr 與 Romer 王不見王，互相都不引用對方的研究。 + +有了這些背景之後，作者認為追逐經濟成長，就是在問怎麼追逐科技發展，和怎麼增加知識與想法的生產。作者認為有四件事情要處理： + +第一是去問誰擁有和控制知識與想法：也就是智慧財產權。智財保護的用意在於提供發明者發明想法的誘因。保護得太短則沒人想發明，保護得太長知識又無法流通發揮影響。作者指出目前的智財制度有老舊（保護太長）、氾濫（保護濫發）、和武器化（用來做反競爭用途）的問題。 + +第二是要增加研發的投資。目前全世界平均對研發的投資是 GDP 的 1%~4% ，其中前三名：以色列大概 6%，韓國 5% ，台灣 4% 。國家和大企業比較不公平，但大科技公司的研發佔獲利比： Alphabet 15%, MSFT 13%, FB 21% 。作者是認為國家應該要再把研發的比例拉更高。 + +第三是要讓更多人們參與研發。研究顯示低收入的人們在長大比較不容易成為發明家。如果這個問題可以緩解，一方面可以解決因收入造成的機會不平等，另一方面更多人參與研發可以促進成長。 + +第四是要增加研究員的效率。我們研究的效率降低了，最低的果實似乎摘完了： + +- 摩爾定律的確從 1970 至今穩定發揮，但要達成同樣的效果，今天要比 1970 花 18 倍的人力。 +- 發現上帝粒子用了 5000 個研究員，33 頁的論文只有九頁是真正內容，剩下的頁面都在列出作者名稱。 +- 大學的行政人員膨脹：哈佛有 7000 大學部學生，但有 7000 行政人員。 +- 研究員花 40% 時間在申請經費。 +- 近期調查顯示 80% 研究員在受申請經費條件的限制之下，做不喜歡的研究主題。 + +作者提到，要改善研究效率，可能沒辦法純靠人了。以蛋白質折疊問題而言，以往研究一個蛋白質的立體形狀幾乎就是用掉一個 PhD 的研究生涯，在今天也只有 17% 已知的蛋白質的立體形狀。但 2022 DeepMind 的 AI 一口氣就發現一億種蛋白質立體形狀了。 + +不過書裡最有趣的是，作者認為哪些作為是無法提升經濟成長太多的。 + +- 造橋鋪路，蓋更多基礎建設：大型建設造成很多金錢浪費。而且這個思路陷入資本導向的思維。在成長模型中，資本增加的邊際效用會遞減。所以理論上不可行。 +- 更好的土地利用與都市規劃：Paul Romer 喜歡都市，因為他認為這是大量的人能在同一個地方見面，促進知識交流，而知識就是經濟成長的養分。但作者認為沒有任何實證資料佐證，在現在遠端工作科技盛行之下，實體見面真的那麼重要嗎？ +- 增加教育：很少國家能讓 90% 以上的人上完初中，50% 的人上完大學。作者認為過去「人力資本的世紀」已經無法再發生了。我認為這是作者最有趣的觀點。我本來也覺得教育的改良應該有很多潛力。但即使我們有能力把人的能力再提升兩倍？或甚至十倍。能比得上蛋白質立體形狀那種 AI 的生產力進步嗎？ + +## 成長的方向 + +作者指出，人們很容易把經濟成長想成一台火車頭。成長的人認為要加速，棄成長的認為要減速或倒車。但這個比喻的問題在於把成長看成像鐵軌一樣只有一個方向，只能前進或是後退。 + +比較好的比喻是像汪洋中的一條船，船有不同的方向可以前進，而我們在看成長時也要選擇往哪個方向前進。 + +一個重要的觀察是：同樣在英國 5% 的成長，現在的 5% 與和在一百年前的 5% 生產出來的東西是非常不一樣的。 + +這當中的差別當然是我們現在有完全不一樣的科技。但科技的發展難道是我們有辦法控制的嗎？ + +作者區分了兩種概念：誘發性的科技發展和引導性的科技發展。 + +誘發性的科技發展的例子像是英國工業革命。為什麼同樣時期，鄰近的德國法國和比利時地區的人沒發展出工業革命？那是因為當時的英國同時有高工資和低燃料成本的環境，使得投入蒸汽機的改良去取代勞工有利可圖。類似的例子還有：在日本照護機器人很先進，因為老人人口多，移民又很低。以及中國的工業機器人，因為製造業以往倚賴的低工資環境已經不在。 + +人們近年對引導性的科技發展的理解大幅增加。主要是 [[Daron Acemoglu]] 的研究。主要論點大概是權力與進步這本書的內容。大致而言，那些誘發科技發展的因素或誘因，並非我們完全沒有能力控制的。透過稅與補助、法規、與社會規範的改變，我們有能力引導科技發展的方向。 +- 疫情期間的隔離政策可以視為，公司要聘人進辦公室的工資成本無限大，因此誘發了 Zoom skype 等遠端工作的科技與文化發生。因為這個政策的因素，催生的這些科技的進程。 +- 疫情初期，疫苗本來預期要十年才能開發出來。但我們八個月之後就有 AZ 和 BNT 。幾個月之後就有莫德納和嬌生。這也是工資成本無限大促發疫苗開發的強烈誘因。 +- 壞的方面是，很多公司因應疫情，把那些需要見面容易傳染的工作自動化了。這造成很多勞工的困境。 + +引導性科技發展的理論其實受到許多左派右派的挑戰。 + +右派如市場基本教義派的如胡佛研究所就會說，引導性科技發展是一種政策干預市場價格機制正常運作的方式。作者的反駁是，對我們作為「消費者」身份時，市場機制會反映我們在乎的事情。但我們作為「公民」身份時，市場機制難以反映我們的顧慮。作者也回應引導性科技發展並非中央規劃。 + +左派則批評引導性科技發展仍然是利用財產權、誘因、市場機制去鞏固經濟成長和資本主義，像是請鬼拿藥單。作者則是提醒他們忘記左派文獻的根源：經濟成長並不是資本主義獨有的東西。在以前社會主義社會，他們重視成長更甚於資本主義社會，並以物質產出的數量為傲。 + +作者引用 Derek Parfit ：我們目前處於一個歷史的轉捩點。在過去兩百年間科技高速變化。我們會有更大的能力去改變環境、甚至是自己或後代。如果我們在接下來幾百年做明智的行動，人類可以度過這個最危險最關鍵的時期。 + +作者說，如果我們有更多能力去改變未來，那代表過去能給我們的經驗就變少了。例如：以前的人批評「綠色成長（Green Growth）」不可能發生，但以現在的資訊來看，這在未來極有可能。 + +如果科技發展真的是我們有辦法引導其方向，這也代表我們對我們選擇的方向有道德責任。 + +## 成長的取捨與道德責任 + +前面提到，經濟成長有氣候的衝擊，不平等的增加，工作受到威脅等等代價。我們要如何在眾多重視的價值中做取捨？ + +首先作者先討論，取捨是必要的嗎？也許我們可以先想，有沒有避免抵換的方式？（Pareto 改善）再者，如果完全避免是不可能的，那有沒有弱化抵換的方式？窮盡這些選項之後，我們可以再來談取捨。 + +- 避免抵換 + - 前面提到如果我們給低收入的人更多機會，這能夠同時改善不平等和增進成長。 +- 弱化抵換 + - 作者花了很多力氣談綠色成長：也就是在不增加環境負擔之下增加經濟成長。 + - 2007 Stern review: 要把 CO2 降到現在的 80% 以下，要花費每年 1% GDP 。 + - 但到了2020 Climate Change Committee。完全消滅排碳的成本只需要 0.5% GDP 。 + - 1976 太陽能價格是每瓦 100 美元。到 2019 變成每瓦 0.5 美元。 + - IEA: 對 2030 年全球使用太陽能的預測。 2021 比 2006 年的預測多 30 倍。預期在 2027 年太陽能超越煤炭和天然氣。 + - 綠色脫鉤：許多國家正在發生 GDP 成長又同時減少排碳的情況。 + - 綠色脫鉤可以說是發明家、企業家等集體努力之下而成。透過文化、制度、稅與補助、法規、及社會敘事與規範共同塑造。 + - 烏俄戰爭在俄斷天然氣之下，間接造成歐洲綠能轉型。 + - 自動化 + - 比爾蓋茲觀察到：為什麼工人要繳稅，但工廠把工人換成機器人之後，機器人不用繳稅。這裡反映的是，勞力和資本同樣是生產要素，但勞力太容易被課稅。 + - 讀者可能會很想對資本課稅，來減少成長對不平等的抵換。但對資本課稅要很小心，一不小心容易得到反效果。 + +談完避免和弱化抵換之後，便能來談接受抵換的情況。 + +要談成長與環境、不平等等抵換，其實是在討論平衡現在人與未來人的利益的問題。現在成長不足，是留給未來比較少的手段解決問題，但也不能留給未來人一些不平等和污染的爛攤子。 + +經濟學裡面有個「代際平等（intergeneration equity）」的題目。通常是用折現率這樣一個參數，來表達我們怎麼平衡現在人與未來人的利益。 + +作者認為，現在的人與文化的確是短視近利。企業極大化短期報酬來取悅投資人，政府追逐短期成就來取悅選民，而民眾則是被每天的煩惱所困擾。這算是一種極重視短期的極端。 + +作為另一個極端的比較，作者提到矽谷最流行的長期主義。（他說是 exteremely helpful mistake）。 + +長期主義關注幾百萬年、幾千萬年、或幾億年以後。人類已經突破 80 兆人口，在那個最遠的數字，地球已經被燃燒殆盡的太陽給吞噬，人們已經在外太空飄了。 + +長期主義者認為，我們應該要為那些遠在未來的人謀福利，因為他們人數比我們大太多了。 + +但作者說長期主義者想保護的未來太遙遠了，而且是我們無法想像。我們有可能近期就長壽運動成功，突破長壽逃逸速度（ longevity escape velocity），也有可能上傳我們的心智達到數位永生。但這些事情都已經超越我們的想像，更何況那些幾千年幾億年的東西。 + +作者引用 Regina Rini ：「我們試圖預測數百萬年後星際旅行者的需求，就如同舊石器時代的酋長為二十一世紀的矛頭準備尖銳燧石一樣不切實際。」 + +作者還偷酸了一下長期主義的重要支持者 SBF：「結果人類最致命的威脅不是未來的核戰或超級人工智慧，而是在我們時代的一個三十歲捲捲頭髮的億萬富翁」。其經營的 FTX 交易所糟糕的內控造成災難一場。 + +因為未來科技發展的難以想像，作者認為我們應該擁抱「想像力的謙遜」及「務實的謙遜」，去看待未來對現在的需求。 + +所以我們欠未來人什麼呢？作者認為我們應該要比現代人再更加關心未來，但比長期主義者關注再更加近的事情。不能和經濟學家一樣爭論折現率的幾個小數點，但必須爭論各種議題的取捨方向。 + +對於道德問題，作者最後還是丟了「要讓公民決定」這種答案。作者提了一些類似國民法官這種 Mini-publics 的機制，在國家中抽樣人民去為特定的議題找答案。 + +## 結論 + +作者基本上談論經濟成長的起源，人們做了哪些研究，並討論經濟成長的好處與代價。在平衡好處與代價之間，並非如現在一昧追逐成長和完全放棄成長這麼極端，而是有許多方向與中間地帶可以取捨。 + +我覺得這本書是我最近看了幾樣東西的交匯點。 + +- 我可以理解為什麼 Glen Weyl 要那麼強調報酬遞增的研究（現在諾貝爾獎頒太多給報酬遞減的研究）。 +- 知識、想法在經濟上的重要性。 +- 權力與進步講的引導性科技發展為什麼重要。 +- 我們看到各種新科技要怎麼放到經濟成長的角度去理解。 + +另外也看到很多之前沒關注到，重要的事。例如綠能的發展，蛋白質的摺疊。 diff --git a/content/posts/2025-10-03_vibe_coding2.zh-TW.md b/content/posts/2025-10-03_vibe_coding2.zh-TW.md new file mode 100644 index 0000000..3dc9453 --- /dev/null +++ b/content/posts/2025-10-03_vibe_coding2.zh-TW.md @@ -0,0 +1,52 @@ +--- +title: "隨興開發聽起來很瞎" +date: 2025-10-03T00:00:00+08:00 +draft: false +--- + +隨興開發（Vibe coding）聽起來很瞎。因為遊戲規則是開發者不能去看或動 AI 寫出來的程式。 + +也許那是我們才剛穿越到 2025 年，還不習慣這種做法。寫過程式的人，會在過去的經驗中必須了解程式碼的所有細節。或，真的嗎？ + +我發現我從來沒有管過編譯器下面的事。假設用 Python 寫了程式，看起來沒問題，執行起來沒什麼問題，我就覺得沒問題了，非常 Vibe。 + +一種原因是程式語言編譯後的錯誤，通常很罕見也很安靜，所以人們不怎麼檢查，出事也不怎麼嚴重。 + +直接問 AI 有什麼上過頭條的編譯器錯誤，他能給出最接近的答案是 1994 的 Pentium FDIV bug ，和編譯器無關。這個事件是數學系教授在研究數學時，發現有個除法一直算錯，追究下去發現是晶片的問題。 + +再對 AI 嚴加逼問之後，發現最近最有名的編譯器錯誤就發生在我們區塊鏈圈子。 2023 年，撰寫合約用的 Vyper 語言，因為編譯器的實作錯誤，用來避免重入攻擊的函式失去效果。駭客因此漏洞從 Defi 專案的合約竊取 7000 萬美金。 + +且不說編譯器這麼遙遠的主題。多少工程師是有辦法檢查每個專案依賴的套件？我能做到的就是看起來可疑的套件不要用。 + +或整個軟體開發上，為了減少開發者認知負擔，本身就需要重重抽象。套件是別人包好的，開發者可以在不需要知道背後的運作細節的情況使用它。函式也是，當人們寫一個 get_post_by_id ，不用知道背後的實作細解也大概可以想像這段程式的用途。 + +所以本來開發上就有某種程度的瞎，與適度的瞎。 + +瞎的部分是問題會發生的所在。想辦法補起來就好，但不一定總得靠人力。 + +從編譯器的思路出發，隨興開發就好像某種更高階的語言一樣。以往的程式語言，是模擬英文的感覺，但加上非常嚴格的語法，讓人們可以和電腦溝通。高階的程式語言需要經過層層翻譯，先變成比較低階的語言，再變成組合語言，再變成機器碼，叫石頭與磚頭開始工作。 + +現在則是我們用非常不精確的人話溝通。 AI 聽得懂，並翻譯成以往的程式語言。這整串流程只是多加了一個 AI 翻譯官。 + +1940 年代，人們最早是透過機器碼和組合語言和機器溝通的。早期的高階語言可以拉到 Fortran (1957) 或 LISP (1958) 。 + +最早的時候機器超貴，造價是年薪 100 倍（ 1950 機器 100 萬美金，工程師年薪 1 萬美金），但現在工程師的年薪可以買 50~100 台筆電（2025 筆電 2000 美金，工程師年薪 10 萬美金）。因此早期的時候，可以花很多工時，用最接近機器的語言和機器溝通。現代工程師的時間比較貴，用人話才越有效率。 +但人話又可以讓開發這件事情讓非工程師的人勝任。 + +Claude Code 開發者的專訪中，他說人們總說隨興開發讓從零到一這件事情變簡單了。他個人覺得是專案要上線前，往往需要放棄的那 2% 功能，現在可以不用倚賴工程師完成。 + +隨興開發在重塑人與機器的溝通，也在重塑人與人之間的合作方式。或也許單個人本身的合作方式也變了。 + +產品經理與開發者一直有某種利益衝突的關係。前者需要更多的功能上線，但後者需要為開發新功能帶來的後果負責，那包含開發的耗時與耗神、上線後的故障等。如果一直推了新功能然後又撤掉，開發團隊士氣會燃燒殆盡。 + +我自己寫程式的時候，一直要把想要的新功能與自己的耗時和疲勞權衡。但叫 AI 寫程式的時候，我可以一直推功能，一直欠技術債。產品經理腦不會和開發者腦打架。把開發的工作推給 AI 解決了那種利益衝突。 + +難道這代表隨興開發之下，真的不再用人眼去看以往的程式語言了嗎？不盡然。在高階語言開發之下，會需要動到低階語言的先例不是沒有。 + +人們常會在效能熱點的地方，換成人比較不好理解但高效能的低階語言實作。在區塊鏈圈子看到的是 Defi 或密碼學的數學會用省油的組合語言實作。低階語言效能高的原因其實是高階語言在翻譯下來的過程中，編譯器往往會以保守安全並犧牲效能的方式翻譯。但如果人們在熟悉業務邏輯的情況下撰寫低階語言，可以知道哪些不必要的步驟可以省略，達成節省運算的目的。 + +大約一年以前，我需要整合某個新密碼學套件，做某種面向使用者的應用。但因為當時 AI 對一個全新套件以及其密碼學並無理解。我就再把密碼學套件再包成數個符合業務邏輯的函式， AI 就能接手後面的工作了。用函式包裝程式的作法在和人協作的時候很標準，但在和 AI 協作時，好像是為了某種高階語言去用低階語言去最佳化的感覺。 + +我們的意念變成機器的行動的距離變近了。開發變得像《猴爪》小說一樣，要小心自己許的願望，並一直調整自己許的願望是什麼。 + +這樣想想就覺得還要開口和寫訊息和機器溝通好冗贅。我們需要魔法。 diff --git a/content/posts/2025-10-07_lost_colony.md b/content/posts/2025-10-07_lost_colony.md new file mode 100644 index 0000000..ab787eb --- /dev/null +++ b/content/posts/2025-10-07_lost_colony.md @@ -0,0 +1,79 @@ +--- +title: "Review: Lost Colony by Tonio Andrade" +date: 2025-10-07 +draft: false +--- + +{{< translation type="machine" from="Chinese" />}} + +![](https://books.google.com/books/content?id=wGmYDwAAQBAJ&printsec=frontcover&img=1&zoom=1&edge=curl&source=gbs_api) + +I finished reading Tonio Andrade's [[Lost Colony: The Untold Story of China's First Great Victory over the West]](https://www.eslite.com/product/1001110932643195). The original was published in 2011, with the Chinese edition republished in 2017. Many Dutch names in it appear closer to colloquial pronunciations. + +This book mainly tells the story of Koxinga's 1661 attack on the Dutch at Fort Zeelandia. Why did the author want to research this war? It turns out there's a major world history question to answer. + +Why did relatively backward Western European countries come to dominate the globe in the 16th century? Traditional historians have proposed many different reasons: guns, property rights, economics, politics, social institutions, etc. But for every reason traditional historians propose, global historians counter with examples of others who had those factors but didn't dominate the globe. + +Global historians believe the Europe-Asia divergence wasn't in 1492 when Columbus sailed, 1497 when da Gama rounded Africa, or 1600 when the English and Dutch East India Companies were established. They believe the so-called Great Divergence happened with industrialization in 1800, when Europe truly became decisively powerful. + +One theory historians debate about Europe-Asia power is the "Military Revolution thesis." Frequent European wars created innovation: powerful guns, ships, troops, fortresses—these military innovations then sparked social institutional innovation. + +The global history school calls itself historical revisionism. Both they and their opponents accept the Military Revolution thesis but reach different conclusions. Revisionists believe Europe indeed had some advantages, but not many. Anti-revisionists believe Europe became strong precisely because of the military revolution. + +The 1661 war is a battlefield for debating the Military Revolution thesis. Anti-revisionists believe the Dutch occupied Taiwan due to their powerful technological strength. + +Revisionists believe the Dutch could occupy Taiwan because China, Japan, and Korea let them, as everyone was busy—the Dutch seized the opportunity in a naval power vacuum. When the Ming needed Taiwan, they took it right away. + +Tonio Andrade is a revisionist. Before writing Lost Colony, his previous book argued how easily Koxinga defeated the Dutch. But after examining Chinese and Dutch historical sources, he found this war wasn't as simple as he'd imagined. + +Guns, the anti-revisionists' favorite, weren't simply a European advantage. The Ming had long mastered musket technology but found it impractical and chose not to use it. + +The Ming, founded in 1368, was a gunpowder empire. They used cannons against neighbors, who then learned artillery. Those neighbors used artillery against other neighbors. Thus artillery spread westward. + +Muskets fired slowly, so different peoples independently evolved volley fire tactics. The first rank fires then the second, while the first reloads behind, using multiple ranks to reduce reloading gaps. This volley fire demanded discipline to avoid breaking in chaotic battlefields. + +After learning about muskets, the Ming found them impractical. Though accurate, efficiency dropped against human wave tactics. So Ming armies kept giant cannons but didn't use guns. + +Koxinga's army bypassed Dutch defenses through a breach in the Luermen sandbar, camping in northern Tainan. They quickly besieged Fort Provintia (today's Chihkan Tower), forcing Jacob Valentine to surrender. Frederick Coyett, the Dutch governor at Fort Zeelandia, sent Thomas Pedel with 240 musketeers against Koxinga's thousands. + +Koxinga deployed his elite Iron Men troops wearing iron armor and masks. Pedel had previously used hundreds of musketeers against six thousand in the Guo Huaiyi Rebellion without injury, so he underestimated this battle. When Koxinga's Iron Men were shot, replacements immediately filled gaps, maintaining formation. Pedel fell into a rear ambush, was routed, and died on the spot. + +Dutch muskets didn't defeat Koxinga's army. Musketeer discipline wasn't superior to Koxinga's troops' discipline. + +Andrade believes the Dutch indeed had special technology, and if Coyett hadn't made a series of wrong decisions, they might have defeated Koxinga. + +The Dutch's main advantages were Renaissance star forts and ships. + +The Koxinga facing Coyett was already a siege veteran of dozens of battles. Chinese cities are enormous, with walls wide enough for carriages. Fort Zeelandia was tiny compared to even rural Chinese fortresses. Yet it held Koxinga off for a year. + +This was mainly because Renaissance star forts evolved for the artillery age. The fort could create crossfire, with artillery firing from different directions at any approach. Koxinga, raised on Sun Tzu's Art of War, considered sieges the worst strategy, so he first wrote many letters urging Dutch surrender. When he actually needed to siege, he threw massive manpower at it. Chinese walls quickly fell once breaches were found and packed with gunpowder. + +Against star forts, one must build siege works nearby. Finally, a German drunkard who defected from the Dutch army taught Koxinga's forces this trick, prompting Coyett's surrender. + +For ships, Dutch advantages were sailing against the wind and broadside cannons. Sailing against the wind let Dutch ships sail to Batavia for help when monsoon winds were unfavorable. Broadside cannons increased ships' firepower output. So in Zheng Zhilong's era, defeating Dutch ships required luring them deep inland then attacking with fire ships. + +Beyond world history's big questions, this book's narrative is brilliant, with vivid character portrayals. + +Coyett was basically a corporate drone who held meetings and left paper trails before doing anything. He worked meticulously, giving detailed instructions and leaving contingency plans. + +His problem was discord with colleagues. When he initially anticipated possible invasion by Koxinga's forces, Batavia had actually reinforced defenses, but he drove away his colleague Van der Laan who led elite naval forces. When reinforcements under Jacob Caeuw finally arrived after holding out, he didn't trust them, leading to a disastrous counterattack. The Batavia governor also didn't trust him, leading to his eventual investigation, exile, and disgrace. + +Koxinga had strategies trusting Dutch military defectors, but Coyett didn't trust the Chinese farmer defector Su. The author believes following Su's suggestion for naval blockade of Koxinga's supplies could indeed have starved Koxinga's forces. + +Finally, He Bin. Why did Koxinga, while successfully fighting to restore the Ming and already at Nanjing's gates, suddenly attack Taiwan? Because He Bin told Koxinga that Taiwan had abundant provisions. When Koxinga arrived in Taiwan and found nothing, he had to order soldiers to farm. + +He Bin's father served as Dutch interpreter and owned extensive lands. + +At the time, businesspeople couldn't trust merchants' scales, fearing false weights, so many places established public scales. He Bin proposed building public scales to the Dutch. + +The Dutch thought scale positions should be auctioned. Three Chinese bid, but He Bin somehow always got the position. The Dutch tax official was furious. Angry merchants sued He Bin, making his position untenable. + +At the time, Koxinga implemented naval embargo to obtain food for counterattacking the Qing. Coyett, newly in position, wanted to restore trade and communicate with Koxinga. He Bin intermediated, telling the Dutch everything was arranged with Koxinga and the embargo would lift, while telling Koxinga the Dutch would pay tribute. + +My impression reading this is that Taiwan keeps getting sold out by political-business brokers. And Koxinga's invasion based on false provisions intelligence is an example of Chinese miscalculation in invading Taiwan. + +Finally, I don't know where to place this, but the book has a beautiful passage describing the Little Ice Age, here are excerpts: Spanish soldiers heard Philippine volcanoes erupting; astronomers in Korea, China, and Europe recorded sunspots; tree rings in northern Italy tightened from lowered temperatures, making seventeenth-century violins sound special. Mexico had no rain, the Nile dropped to its lowest level. + +![](images/waaggebouw.jpg) + +Image: Dutch using public scales and collecting goods tax from children at Fort Zeelandia, Tainan, Taiwan. diff --git a/content/posts/2025-10-07_lost_colony.zh-TW.md b/content/posts/2025-10-07_lost_colony.zh-TW.md new file mode 100644 index 0000000..c1d2073 --- /dev/null +++ b/content/posts/2025-10-07_lost_colony.zh-TW.md @@ -0,0 +1,76 @@ +--- +title: "《決戰熱蘭遮》閱讀心得" +date: 2025-10-07 +draft: false +--- + +![](http://books.google.com/books/content?id=cscJEQAAQBAJ&printsec=frontcover&img=1&zoom=1&edge=curl&source=gbs_api) + +看完了歐陽泰的【[決戰熱蘭遮](https://www.eslite.com/product/1001110932643195)】。這本原文書是 2011 出版，中文 2017 再版，裡面很多荷蘭人的名字看起來比較貼近民間唸法。 + +這本主要在講 1661 鄭成功攻打熱蘭遮堡荷蘭人的故事。為什麼作者想要研究這場戰爭？原來是有世界史上的大哉問要回答。 + +為什麼相對亞洲落後的西歐國家，會在 16 世紀稱霸全球？傳統的史學家提了很多不同的原因：槍炮、財產權、經濟、政治、社會制度等。但傳統史家每提一個原因，全球史學者就會提出反例，說誰誰誰也有那些因素，為什麼他們沒稱霸全球。 + +全球史學者認為歐亞實力的分歧不在 1492 哥倫布啟航 、1497 達伽馬繞過非洲 、或 1600 英國和荷蘭東印度設立。他們認為在 1800 工業化才發生所謂的大分流，歐洲真正決定性的強盛。 + +其中史學家爭論歐亞實力的一個理論是「軍事革命論」。歐洲國家頻繁的戰爭造就創新：強大的槍炮、船艦、部隊、堡壘，這些軍事創新又連帶開啟社會制度的創新。 +全球史學者這派稱自己是歷史修正主義。他們與反方都承認軍事革命論，但各自給出不同的結論。修正主義認為歐洲的確有些許優勢，但沒太多。反修正主義者則認為歐洲就是因為軍事革命才強。 + +對於 1661 的這場戰爭，就是軍事革命論的一個辯論戰場。反修正主義者認為，荷蘭人因為其強大的科技實力，才佔領了台灣。 + +修正主義者則認為，荷蘭人可以佔領台灣，是因為中國、日本、朝鮮讓荷蘭人佔，因為彼此都很忙，海權真空之下荷蘭人才有機可趁。明朝人需要台灣了，一下子就把台灣拿下來了。 + +歐陽泰是修正主義者。在寫決戰熱蘭遮之前，他的上一本書就是在講鄭成功打荷蘭有多輕鬆。但後來他去翻閱中荷雙方的史料之後，發現這場戰爭沒他想像的那麼單純。 + +反修正主義者最喜歡的槍炮，並非單純是歐洲人的優勢。明朝人早已掌握火槍的技術，但覺得不好用，選擇不用。 + +明朝在 1368 創立，是一個火藥帝國。他們用大炮打鄰國，鄰國因此學會火砲。鄰國再用火砲打其他鄰國。因此火炮因此一路傳到西方。 + +火槍的射速很慢，因此不同民族都獨立演化出排槍戰法。第一排射完換第二排射，第一排到背後裝子彈，以多排火力去減少換子彈的空檔。這種排槍戰法很講究紀律，在面臨混亂的戰場不能潰散。 + +明朝人學到火槍之後，覺得火槍不太好用。雖然可以射得很準，但被人海戰術一壓效率不高。因此明朝軍隊留著巨型的大砲，但不使用槍。 + +鄭成功的軍隊在鹿耳門沙洲的破口繞過荷蘭的防線，在台南北部扎營。一下子就圍困普羅民遮城（今赤崁樓），促使貓難實叮（Jacob Valentine）投降。駐守熱蘭遮堡的荷蘭長官揆一（Frederick Coyet），派遣拔鬼仔（Thomas Pedel）率 240 火槍兵對鄭成功數千人。 + +鄭成功派了身著鐵甲面具的虎面精銳部隊抵擋。拔鬼仔曾經在郭懷一事件用數百人火槍兵抵六千人，無傷通關，因此在此役中輕敵。鄭成功的虎面精銳部隊一被射中之後，馬上補人上去，維護住陣型。拔鬼仔又中了背面埋伏，因此潰散大敗，命喪當場。 + +荷蘭人的火槍並沒有打贏鄭成功的軍隊。因為火槍兵的嚴格紀律也並沒有比鄭成功軍隊的紀律厲害。 + +歐陽泰認為，荷蘭人的確有很特別的科技，而如果揆一沒有做出一系列錯誤決策，其實有機會打贏鄭成功。 + +荷蘭人主要的優勢在於文藝復興的星狀堡壘與船隻。 + +在面對揆一時的鄭成功，已經是打過數十場圍城戰的攻城老手。中國的城都十分巨大，城牆上面可以行駛馬車。熱蘭遮堡比起來是個比中國鄉下都還要小很多的堡壘。但可以整整卡鄭成功一年。 + +主要是 文藝復興的星狀堡壘是炮火時代下演化出的產物。城堡可以形成交叉火網，從不同的方向進攻都會被火砲射擊。鄭成功是學孫子兵法長大的，認為攻城是最下策，因此都寫了一堆信先勸降荷蘭人。他實際需要攻城的時候，攻城也是大量人力壓上去，中國的城牆只要找到缺口埋火藥，很快就破了。 + +對付星狀堡壘，必須要在附近構築攻城工事。最後是荷蘭軍叛逃的日耳曼酒鬼教會鄭軍這招，才促使揆一投降。 + +船的部分，荷蘭的優勢是逆風航行的能力和船側舷炮。逆風航行的能力讓荷蘭的船隻，可以在季風風向不順時，駛往巴達維亞求援。船側舷炮讓船隻能輸出的火力上升。因此鄭芝龍時代要對付荷蘭船隻，都必須用計引誘荷蘭船隻深入，再用火燒船去攻擊。 + +除了世界史的大哉問。這本書的敘事也很精彩，每個角色的形象都很生動。 + +揆一基本上就是個社畜，做什麼事情會先開會，留下書面記錄（ paper trail ）。他做事很細心，吩咐下去的事情很詳細，還會留下應急的錦囊妙計。 + +他的問題就是和同事不合，一開始預期鄭軍可能入侵的可能，其實巴達維亞有加強防禦，結果他把率領精銳海軍的同事范德蘭氣走。好不容易堅守到援軍卡烏來了，但不信任援軍，最後導致一場災難性的反攻。巴達維亞長官也不信任他，導致他最後被調查和放逐，身敗名裂。 + +鄭成功有相信叛逃的荷軍的策略，但揆一沒相信叛逃的漢人農民蘇。作者認為如果有照蘇的建議海上封鎖鄭軍的補給，確實有機會餓死鄭軍。 + +最後是何斌。為什麼鄭成功反清復明好好的，都打到了南京城，突然要打台灣？就是何斌去和鄭成功說台灣有一堆糧草可以取得。結果鄭成功來到台灣時發現什麼都沒有，必須要叫士兵去屯墾。 + +何斌的父親擔任荷蘭通譯，擁有眾多田地。 + +當時做生意人們沒辦法相信商人的秤，怕他們虛報重量，因此很多地方會設立市秤。何斌向荷蘭人提議蓋市秤。 + +荷蘭人認為市秤職位要拍賣，有三個漢人來標，但何斌不知道為什麼總是最後取得職位。荷蘭稅務官非常生氣。一群憤怒的商人怒告何斌，讓何斌快混不下去。 + +當時鄭成功為了反攻清朝，要取得糧食，因此在海上實施禁運。揆一剛接手職位，想要恢復貿易，並和鄭成功溝通。何斌當兩個人的中間人，一方面跟荷蘭人說都和鄭成功談好了，會解除禁運，另一方面和鄭成功說荷蘭人會納貢。 + +我看到這邊的感想是台灣都是被一些政商掮客給賣掉的。然後鄭成功因為錯誤的糧草情報侵台，是一種中國誤判侵台的例子。 + +最後我不知道應該擺在哪段，這本書有一整段最美的小冰期的描寫，節錄幾句。西班牙士兵聽見菲律賓的火山爆發；朝鮮中國歐洲的天文學家記錄到了太陽黑子；義大利北方的樹木年輪因為降低的溫度變得緊密，導致十七世紀的小提琴音色特別。墨西哥沒有下雨，尼羅河水位降到最低。 + +![](images/waaggebouw.jpg) + +圖：荷蘭人使用市秤和小朋友收受貨物稅示意圖 diff --git a/content/posts/2025-10-07_lost_colony/images/waaggebouw.jpg b/content/posts/2025-10-07_lost_colony/images/waaggebouw.jpg new file mode 100644 index 0000000..4c66408 Binary files /dev/null and b/content/posts/2025-10-07_lost_colony/images/waaggebouw.jpg differ diff --git a/content/posts/2025-11-05_triumph_of_the_city.md b/content/posts/2025-11-05_triumph_of_the_city.md new file mode 100644 index 0000000..6249863 --- /dev/null +++ b/content/posts/2025-11-05_triumph_of_the_city.md @@ -0,0 +1,71 @@ +--- +title: "Review: Triumph of the City" +date: 2025-11-05 +draft: false +--- + +{{< translation type="machine" from="Chinese" />}} + + +![](https://books.google.com/books/content?id=-yWTIKsWGm4C&printsec=frontcover&img=1&zoom=1&edge=curl&source=gbs_api) + +"Triumph of the City" is a great book. I've always wanted to read something about cities, and this book was very satisfying. The author is Edward Glaeser, one of the advisors of [Jheng, Shao-Yu](https://sites.google.com/view/jheng-shao-yu), who is a cool scholar I've been following. + +This is an accessible book—reading it feels like taking trips to many cities at once, as the book introduces the history and scenery of many cities. It's also a book that builds thinking frameworks and is full of opinions. + +There are no formulas or overly difficult concepts in the book; anyone who has lived in a city can relate to the text. But there are many numbers for evaluating decisions, like "one additional year of education increases salary by 8%." The book also contains many counterintuitive insights. + +My digested understanding is this: the urban issue has only one parameter—population density. + +Places with high population density are called cities, low density places are suburbs. + +High population density has many benefits: + +1. People's talents spark off each other across short distances. Human capital has positive externalities, because working with excellent people teaches you a lot and makes you excellent too. +2. In high-density places, niche or specialized industries can more easily find consumers to share expensive fixed costs. For example, theaters developed in London because there were enough people to watch. + +But high population density also has downsides. These downsides are more like problems that can be solved. + +First, disease and filth become serious with population concentration. So we need to find ways to bring clean water in and drain sewage out. There must be corresponding infrastructure to handle these problems. + +Also, population concentration causes traffic congestion, but building more roads is useless—it just brings more cars to jam together. The author believes in congestion pricing. + +The first thing people might know about Vickrey is "second-price auctions"—the principle being that winners should pay for the losses their participation causes others, namely the second-highest bidder's willingness to pay. Vickrey once advocated congestion pricing because driving causes congestion for others, and one should pay for that burden. The second thing people might know about Vickrey is that he unfortunately died three days after winning the Nobel Prize—he had a heart attack while driving at night, slumping over the steering wheel. The author thinks he might have been driving at night to avoid traffic. + +Congestion pricing is hard to implement because collecting money from voters is politically unpopular. Citizens would rather endure invisible time losses on the road. + +Population density also brings crime. Theft indeed increases because big cities have more targets to "serve." It's mostly poor robbing poor. But crime currently doesn't seem to have as clear an answer as congestion pricing. Much of the discussion is America-specific, so I didn't read as carefully. + +City centers also have many poor people, some areas forming slums. People see the contrast between very rich urbanites and very poor people and think cities having poor people is bad. But the author says cities don't make people poor; cities attract poor people. And many policies aimed at reducing poverty and helping people escape poverty ultimately lead to more poor people in cities, because everyone comes here to escape poverty. + +And this is good, because poor people coming to cities gives them chances to advance. In rural areas, though wealth gaps seem smaller, it's easy to think "this is my life." If von Neumann and Fermi had stayed in rural areas, they'd never have become excellent scientists. + +So how does population density increase? How do cities form? Most cases are influenced by transportation technology and costs. Older cities might have been water transport centers. + +Cities can expand vertically or horizontally. + +Because of steel frame construction and elevators with safety brakes, people could build skyscrapers, letting more people live in city centers. + +Because cars enable commuting, people also live in city outskirts and drive to downtown for work. + +The author thinks horizontal expansion (urban sprawl) is bad—car commuting has higher carbon emissions than walking and public transport in city centers. + +Living in nature-surrounded suburbs is highly polluting behavior. People wanting to protect the environment should obediently live downtown. + +The author thinks people have made many bad "anti-city" policies, including limiting building heights in cities, subsidizing suburban commuting, or subsidizing home buying over renting. + +People limit building heights for urban aesthetics. But if those spaces were filled with tall buildings, they could provide lots of affordable housing instead of climbing prices. The ideal example is like Shenzhen. + +--- + +With this framework, I reflect on Taiwan's news discussions. + +High housing prices seem to consistently rank among top public complaints. + +People complain about high prices in central Taipei. Others say those complaining about high downtown prices just want to speculate on housing. Why not buy reasonably priced houses in New Taipei or Keelung? + +But according to this book, this encourages horizontal urban sprawl, which is bad. + +Taipei seems to have many old four or five-story apartments, unlike Taichung with its skyscrapers to look up at. + +Taiwan seems reluctant to let evil developers do urban renewal, instead trying to remedy housing supply shortages by building public housing. \ No newline at end of file diff --git a/content/posts/2025-11-05_triumph_of_the_city.zh-TW.md b/content/posts/2025-11-05_triumph_of_the_city.zh-TW.md new file mode 100644 index 0000000..aa604ee --- /dev/null +++ b/content/posts/2025-11-05_triumph_of_the_city.zh-TW.md @@ -0,0 +1,69 @@ +--- +title: "《城市的勝利》閱讀心得" +date: 2025-11-05 +draft: false +--- + +[《城市的勝利》](https://www.sanmin.com.tw/product/index/007177000) 這本書好看。我一直想讀一本和都市相關的書，看這本書非常滿足。作者是 鄭紹鈺 的老師 Edward Glaeser 。 + +這是一本好讀的書，讀這本像是一口氣去了好多都市旅行一樣，書裡介紹了很多都市的歷史和風景。這本書也是建立思考框架，充滿意見的書。 + +書裡沒有任何公式或太難懂的道理，住過都市的人都可以體會裡面的文字。但有很多可以評估決策的數字像是「多受一年的教育，薪水可以增加 8%」這種。書裡也有很多違反直覺的道理。 + +我自己消化後的理解是這樣，都市這個議題就是只有一個參數：人口密度。 + +人口密度高的地方叫都市，低的叫郊區。 + +人口密度高有很多好處： + +1. 人們的才華會在短短的距離內互相激盪。人力資本有正外部性，因為和優秀的人一起工作也會學到很多而變得優秀。 +2. 在人口密度高的地方，冷門或專門的行業也比較找得到消費者去分攤昂貴的固定成本。例如劇院在倫敦的發展就是因為有夠多人看。 + + +但人口密度高也有壞處。但這些壞處比較像是某些可以被解決的問題。 + +首先是疾病和髒污會因為人口聚集變得嚴重。所以要想辦法讓乾淨的水進來，污水要能排出。要有相應的基礎設施建設處理這些問題。 + +另外是人口聚集交通也會阻塞，但蓋更多路是沒用的，只會有更多的車來一起塞。作者認為要徵收塞車稅。 + +人們可能知道關於 Vickrey 的第一件事是「第二價拍賣」，道理是得標的人要為他參與拍賣所造成他人的損失所付費，也就是第二高價者的願付價格。Vickrey 曾經提倡塞車稅，因為自己開車上路會造成他人塞車，理應對其負擔付費。人們可能知道關於 Vickrey 的第二件事是他在獲頒諾貝爾獎的三天後不幸離世，他半夜開車心臟停止，趴在方向盤上。作者認為他可能是為了避免塞車而半夜開車。 + + +塞車稅之所以不容易實行，是因為要和選民收錢錢，政治上沒人敢做。民眾寧願忍受路上無形損耗的時間。 + +人口密集也會帶來犯罪。偷竊類的的確會上升，因為大都市裡面有比較多對象可以「服務」。大多是窮人偷窮人。但犯罪目前好像沒有像塞車稅那麼清楚的解答。裡面很多討論也比較美國相關，我就沒看那麼細。 + +都市的市中心也會有很多窮人，有些地方形成了貧民窟。人們看到很有錢的都市人和很窮的人形成的對比會覺得都市有窮人是一件壞事。但作者說都市並不會讓人變窮，而是都市會吸引窮人。而許多為了減少窮人、幫助窮人脫貧的政策 -- 最後會導致都市裡窮人更多，因為大家都要來這邊脫貧。 + +而這是一件好事，因為窮人來都市才會有機會翻身。在鄉村裡雖然看起來貧富差距不大，但容易「這輩子就這樣了」。如果馮紐曼和費米都留在農村，他們永遠沒機會變成優秀的科學家。 + +那人口密度是怎麼變高的呢？都市是怎麼形成的呢？大多數情況是因為交通和運輸的科技和成本影響的。比較古老的都市可能以前是水路運輸的中心。 + +都市可以垂直擴展也能水平擴展。 + +因為鋼骨架構和有安全煞車的電梯發明，人們可以蓋高樓大廈，讓更多人可以住在都市中心。 + +因為有汽車可以通勤，人們也會往都市外圍住，然後再開車到市中心上班。 + +作者認為水平擴（都市蔓延）是壞的，汽車通勤的碳排放比在都市中心走路和搭公共交通工具還要高。 + +住在被自然環繞的郊區是一種污染嚴重的行為。人們想保護環境，就應該乖乖住到市中心。 + +作者認為人們做了很多「反都市」的壞政策，包含限制都市內的建築高度，補助人們通勤到郊區，或補助人們買房而非租房。 + +人們會為了都市美觀，而限制建築高度。而那些空間如果蓋滿了高樓，就可以提供很多廉價的住宅，而不會房價攀高。理想的例子像深圳那樣。 + +--- + + +有了這樣的架構，我回頭想想之前台灣很多的新聞討論。 + +房價高似乎一直是某種民怨的排行前幾名。 + +人們抱怨台北市中心房價太高。也有人說抱怨市中心房價太高的都是為了想要炒房。明明新北或基隆有些價格適宜的房子為什麼不去買。 + +但如果套本書的說法，這是鼓勵人們往都市邊邊水平發展，是不好的。 + +台北看起來很多很老的四、五樓的公寓，但不像台中有那麼多可以抬頭瞻仰的高樓大廈。 + +台灣似乎不愛讓萬惡建商都更，反而卻要蓋公宅來試圖補救缺少的房屋供給。 diff --git a/content/posts/2025-11-10_general_purpose_tech.md b/content/posts/2025-11-10_general_purpose_tech.md new file mode 100644 index 0000000..7d311f6 --- /dev/null +++ b/content/posts/2025-11-10_general_purpose_tech.md @@ -0,0 +1,104 @@ +--- +title: "Review: Technology and the Rise of Great Powers" +date: 2025-11-10 +draft: false +--- + +{{< translation type="machine" from="Chinese" />}} + + +![](https://books.google.com/books/content?id=tNj2EAAAQBAJ&printsec=frontcover&img=1&zoom=1&edge=none&source=gbs_api) + +International relations experts often think that great powers become great because they develop certain key technologies that strengthen national power. Perhaps productivity increases, or military capabilities become formidable, allowing expansion everywhere. + +So maintaining great power status requires leading in new technologies. But how does one lead technologically? + +Political economy experts might say protecting property rights, freedom of speech, etc., makes it easier for the private sector to develop technology. + +Military determinism experts would say countries with unstable neighbors are more likely to innovate out of fear. Vested interests are less likely to block innovation to maintain their advantages. + +Now with AI technology emerging, the US and China are competing for AI advantage. Which experts should we listen to? + +Jeffrey Ding, author of "[Technology and the Rise of Great Powers](https://press.princeton.edu/books/paperback/9780691260341/technology-and-the-rise-of-great-powers)," offers another perspective: which technology you develop matters. Previous perspectives don't care about which actual technology to develop, treating chips and potato chips equally. + +People have a blind spot when looking at technology—they only see the newest, trendiest breakthroughs. Those technologies capture media headlines and people's attention. The author calls these "leading sectors." + +For these breakthroughs to have impact and improve productivity, they must deepen into various industries. So the author believes we must additionally identify so-called "General Purpose Technologies" (GPT for short). + +Capturing leading sectors can't make you a great power; mature general-purpose technology development can. + +How do we define general-purpose technologies? They must meet three requirements: + +1. Have potential for continuous improvement +2. Can find many different uses across various industries +3. Have strong complementary effects—other industries' technologies need to adapt to them to substantially increase productivity + +Because of these three conditions, GPT "diffusion" to various industries and reflection in productivity numbers takes a long time (usually decades). + +How does the author argue GPTs matter more than leading sectors? By examining history's three industrial revolutions: + +**First Industrial Revolution** + +- International relations mainstream narrative: Leading sector innovations: cotton textiles, iron making, steam power +- General technologies: Iron making, steam engines, factory system + +The problem with leading sectors is that before 1815, those technologies didn't affect overall productivity. The key to Britain's rise was rapidly spreading mechanization everywhere, then mechanization's effects drove productivity growth. + +Steam engines changed processes, spawning mechanical engineering. Production shifted from manual to mechanical. Impact on coal mining promoted iron making, fostering mechanization. + +Why didn't France and Netherlands beat Britain at the same time during the First Industrial Revolution? Knowledge and application disconnected. Developing GPTs requires lots of hands-on, tacit knowledge that can't be explicitly stated. In the 19th century, Britain had over 1,000 industry associations with 200,000 members. France had famous scientists, but Napoleon trained experts for limited political and military purposes, limiting trainees' ability to combine with industry. Netherlands also lacked science-industry integration. + +**Second Industrial Revolution 1870-1914** + +- Leading sectors: Chemicals, electrical equipment, automobiles, steel industry +- General technologies: Chemicalization, electrification, internal combustion engines, interchangeable manufacturing +- Chemicalization: Chemical processes spreading to ceramics, food processing, glass, metallurgy, petroleum and other industries +- Electrification: Central steam engine belt drives → electric motors driving individual machine units + +Germany initially made important chemical breakthroughs. But America won chemical production advantages. + +The reason was American chemists collaborated with mechanical experts. They developed chemical engineering departments and unit operations (breaking chemical procedures into basic operations). + +Mechanically, America developed mechanical engineering curricula and standardized screw threads. + +**Third Industrial Revolution** + +- Leading sectors: Computer industry, consumer electronics, semiconductor industry +- General technology: Computerization (recording production information with computers) + +Japan pioneered many leading sector technologies, and American international relations experts once worried about Japan's computer technology lead. But ultimately America became the computer power. + +- America: Created computer science departments. 1968 ACM curriculum helped schools organize computer education +- Japan: 1991 report showed Japan relied on outsourcing. Companies couldn't maintain full-time staff + +--- + +GPTs demand high human capital—both deep expertise in single domains and ability to combine across domains. The author illustrates two types of AI talent: the former is what people imagine as AI talent, the latter is what people often overlook. + +In 2014, Baidu prominently poached deep learning star Andrew Ng from Google. In Alibaba's 2019 IPO photo was Yuan Wenkai, a warehouse worker with professional automation management knowledge who improved logistics warehouse sorting capacity to 20,000 orders per hour. + +Developing GPTs also requires establishing disciplines. Those subjects that delay my fellow classmates' graduation and the majors students study turn out to be sources of historically crucial national power 🤯 + +--- + +However, the author also says looking at past industrial revolutions involves hindsight. People struggle to predict the next GPT. The author says if he'd written this book 20 years ago, it might have been entirely about nanotechnology. + +For me, I actually don't care much about AI or who becomes a great power, as long as it's not authoritarian states. I care more about technologies I've participated in: how blockchain and cryptography become useful technologies. + +Computer science has another definition of "general": people can command computers through programs to accomplish desired business logic. Ethereum and programmable cryptography have these properties. + +But this "generality" also seems like a very useless "generality." Often when you want to use blockchain for some application, people ask why not just use a database. Even now, besides digital wallets (based on programmable cryptography), I really struggle to recommend people use blockchain or new cryptography technologies, unless you really need blockchain-appropriate tools. + +It's hard to say these technologies aren't ready—blockchain has developed for over a decade. AI draws a picture or speaks a sentence, and people immediately sense its use. Blockchain's value is very abstract—it's more like a security product. Security products need corresponding threats for adoption. Signal had few installations before the Ukraine war, but downloads surged after fighting began. + +Security products should ideally be like HTTPS, adopted without users knowing. Users don't worry about threats but are already protected. + +Obviously imagining blockchain as a leading sector is difficult, but can it become some "blockchainization" general technology? Let's hammer the blockchain nail with the GPT hammer: + +1. Continuous improvement potential: Seems so! Architecture can improve with single computer performance or cryptography performance improvements. +2. Broad use across industries: If we ask people's expectations for online platforms, everyone wants social media algorithms and food delivery pricing not unilaterally controlled by platforms. I believe every online service we use should have some profit-distribution calculations locked on blockchain. +3. Strong complementarity: Maybe. Many "why not use databases" reasons are that systems' weakest links negate all blockchain benefits. For example: since data recording requires human input, blockchain's fairness and transparency are meaningless with possible malicious input. Therefore all effective blockchain applications accommodate blockchain in design, replacing input with economic or cryptographic mechanisms. + +--- + +I also hope GPT definitions don't become post-hoc recognition but provide pre-development technology guidance. \ No newline at end of file diff --git a/content/posts/2025-11-10_general_purpose_tech.zh-TW.md b/content/posts/2025-11-10_general_purpose_tech.zh-TW.md new file mode 100644 index 0000000..cbd4c3a --- /dev/null +++ b/content/posts/2025-11-10_general_purpose_tech.zh-TW.md @@ -0,0 +1,99 @@ +--- +title: "《科技與大國崛起》閱讀心得" +date: 2025-11-10 +draft: false +--- + +![](https://books.google.com/books/content?id=a29jEQAAQBAJ&printsec=frontcover&img=1&zoom=1&edge=none&source=gbs_api) + +國際關係的專家都會想說，大國之所以是大國，也許是他們發展出了某種關鍵科技，使得國力變強。也許是生產力變高了，也許是軍事能力變厲害了，可以到處征討。 +所以要維持大國的國力，需要能夠要在新科技領先。那要怎麼在科技領先呢？ + +政治經濟學專家可能會說要保障財產權、言論自由之類的，民間比較容易發展科技。 + +軍事決定論的專家會說，一個國家身邊有些不穩定的鄰居，會比較因為恐懼而創新。既得利益者比較不會阻擋創新來維持自己的利益。 + +現在 AI 科技出來了，美中在競爭誰能在 AI 科技取得優勢。要聽哪種專家的呢？ + +[《科技與大國崛起》](https://www.eslite.com/product/10072302132682921824002) 的作者傑佛瑞丁提出了另外一種觀點：發展哪種科技很重要。前面的觀點都沒有在乎實際上要發展哪種科技，對晶片與洋芋片一視同仁。 + +人們在看科技的時候有個盲點，就是只看那些最新最潮的關鍵突破。那些科技搶佔媒體版面和人們的注意力。作者稱之為領頭羊產業。 + +那些關鍵突破要能夠發揮影響力，改善生產力，必須要深化到各個行業。所以作者認為，必須去額外辨識出所謂「通用科技」（General purpose technology 縮寫也是 GPT）的東西。 + +取得領頭羊沒辦法幫你變成大國，通用科技發展成熟才行。 + +通用科技要怎麼定義呢？要符合三個要件： + +1. 要有能持續改進的潛力 +2. 能夠找到很多不同的用途，能用在各種產業 +3. 要有強烈互補效應，其他產業的技術需要配合他改進能大幅提升生產力 + +因為這三個條件，所以通用科技「擴散」到各個產業，並反映到生產力數字的提升，需要長遠的時間（通常是數十年）。 + +作者怎麼論證通用科技比領頭羊重要呢？要去看歷史上的三大工業革命： + +第一次工業革命 + +- 國際關係主流敘事：領頭羊產業創新：棉紡織、製鐵、蒸汽動力 +- 通用技術：製鐵、蒸汽機、工廠制 + +領頭羊的問題在於 1815 年以前，那些科技都沒影響總體生產力。英國崛起的關鍵在迅速把機械化推廣到各個角落，然後機械化的效果推動生產力成長。 + +蒸汽機改變了工序，催生機械工程。製程從手工變成機械。對煤礦的影響促進製鐵業，助長機械化。 + +第一次工業革命同樣的時間點為什麼法國和荷蘭沒贏英國？知識與應用脫節。發展通用科技要有很多動手做，無法明說的內涵知識（tacit knowledge）。19 世紀，英國有一千多個產業協會，有 20 萬會員。法國雖然有一堆有名的科學家，但拿破崙以有限的政治與軍事目的培養專家，限制了受訓者與產業結合的能力。荷蘭也是科學界和產業界沒有結合。 + +第二次工業革命 1870~1914 + +- 領頭羊：化學品、電氣設備、汽車、鋼鐵產業 +- 通用技術：化學化、電氣化、內燃機、可互換製造 +- 化學化：化學製程在陶瓷、食品加工、玻璃、冶金、石油及其他產業的普及 +- 電氣化：中央蒸汽機皮帶驅動 -> 電動機驅動個別機器單元 + +原先德國在化學上取得重要突破。但卻是美國贏得了化學上生產的優勢。 + +原因是美國的化學家有和機械專家合作。發展出了化工系和單元操作（把化學程序分解為一系列基本操作）。 + +機械上，美國發展出了機械工程課程，也做了螺紋標準化。 + +第三次工業革命 + +- 領頭羊：電腦業、消費電子業、半導體業 +- 通用技術：電腦化（把生產資訊用電腦記錄） + +很多領頭羊關鍵技術是日本先做出來，美國的國際關係專家一度擔心日本在電腦科技的領先。但最後仍然是美國變成電腦大國。 + +- 美國：變出電腦科學系。1968 ACM 課程大綱，幫助各院校組織電腦教育 +- 日本：1991 報告，日本仰賴外派外包。公司沒辦法養正職 + +--- +通用科技對人力資本的要求很高，既要能對單一領域有深入的手感，又要能夠跨領域的結合。作者舉例展示兩種 AI 人才：前者是人們想像中的 AI 人才，後者是人們常常忽略的 AI 人才。 + +2014 年，百度從谷歌高調挖走深度學習明星學者吳恩達。2019 年阿里巴巴上市時的照片中有位袁文凱，他是一位理貨員，有專業的自動化管理知識，把物流倉庫的分揀能力提升到每小時兩萬張訂單。 + +要發展通用科技，也需要把學科訂定出來。原來那些會卡同學畢業的科目，和同學念的科系，都是歷史上關鍵國力的來源 🤯 + +--- + +不過作者也說，去看過去的幾場工業革命有點是後見之明。人們很難去預測下一個通用科技是什麼。作者說如果他在 20 年前寫這本書，就可能會整本都在講奈米科技。 + +對我來說，我其實不太關心 AI ，也不太關心誰會變成大國，只要不是獨裁國家就好。我比較關心我過往參與過的技術：區塊鏈、密碼學怎麼變成一樣有用的技術。 + +電腦科學上有另一種「通用」的定義：人們可以透過程式去命令電腦完成想要達成的業務邏輯。以太坊、可程式化密碼學具備這些性質。 + +但同時這些「通用」看起來也是一種很沒用的「通用」。往往你想要在某種應用使用區塊鏈，人們就會說怎麼不用資料庫就好。到現在除了數位皮夾（基於可程式化密碼學）外，我也真的很難推薦人們使用某種區塊鏈或太新密碼學的科技，除非你真的是為了使用區塊鏈而需要相應的工具。 + +這很難說是那些科技還沒準備好，區塊鏈也發展十幾年了。AI 畫一張圖，講一句話，人們就能感受到其用途。區塊鏈帶來的價值非常抽象，他比較像個安全產品。安全產品需要有相應的威脅，人們才會取採用。 Signal 在烏克蘭戰爭前沒什麼人裝，開戰後下載量急劇攀升。 + +安全產品理想上應該要像 HTTPS ，在使用者不知不覺的情況下已經採用。使用者不用擔心自己受到什麼威脅，但已經受到保護。 + +顯然要想像區塊鏈是什麼領頭羊大概有點難，但他有辦法變成什麼「區塊鏈化」之類的通用科技嗎？我們拿著通用科技的錘子來敲敲看區塊鏈釘子： + +1. 有沒有持續改善的潛力：好像有！架構可以隨著單一電腦的性能提升或密碼學性能提升而改善。 +2. 能不能廣泛使用在各個產業：如果去問人們對網路平台的期待，大家會想要社群網站的演算法、外賣平台的定價機制不要被平台單方面掌握。我是認為應該每種我們用的線上服務，都要有一部分影響利益分配的運算是鎖定在區塊鏈上。 +3. 有沒有強烈互補性：可能有。很多「為什麼不用資料庫」的理由是系統裡最弱的一環會抵銷區塊鏈帶來的所有好處。例如：因為欲記載的資料也要人為輸入，區塊鏈帶來的公正透明效果，在惡意輸入的可能情況下毫無意義。因此所有區塊鏈有效的應用會遷就區塊鏈去設計，輸入的部分要換成某種經濟或密碼學機制。 + +--- + +我也希望通用科技的定義不要變成一種事後的認定，而是可以提供事前發展科技的指引。 diff --git a/content/posts/2025-11-21_project_hail_mary.md b/content/posts/2025-11-21_project_hail_mary.md new file mode 100644 index 0000000..6910a8e --- /dev/null +++ b/content/posts/2025-11-21_project_hail_mary.md @@ -0,0 +1,35 @@ +--- +title: "Review: Project Hail Mary" +date: 2025-11-21 +draft: false +--- + +{{< translation type="machine" from="Chinese" />}} + + +![](https://books.google.com/books/content?id=-Ff2DwAAQBAJ&printsec=frontcover&img=1&zoom=1&edge=none&source=gbs_api) + +The movie is coming out in 2026. + +I originally thought I wouldn't read any more sci-fi about astronauts and aliens, but since it was a gift, I read it—and it was very readable. + +The book's content is super easy to spoil, but the movie trailer has already spoiled quite a bit, so this article will discuss the following content and parts that don't affect plot progression: +- Humans encounter aliens +- Humans need to solve problems in outer space +- Humans need to mobilize Earth's entire resources to complete a plan + +At first glance, this is an already exhausted topic—how many novels and movies about humans meeting aliens have there been throughout history? Why read another one in 2020? After finishing it, you really feel Andy Weir truly mobilized all of Earth's resources to save this genre. + +This isn't a book you'd describe as "full of imagination"—rather, readers marvel at how much the author "thought things through." The book doesn't introduce many new sci-fi concepts, but each concept solidly influences the others, making the author's sci-fi world feel tangible. + +The sci-fi black box is thinner than a cell membrane. In the novel, the various challenges the protagonist faces are all like science and engineering problems. The seemingly arbitrary numbers thrown around in the book—length, mass, time, temperature, etc.—readers can calculate them all using formulas (unless it mentions the protagonist needs to calculate for a long time with spreadsheets). Readers can also challenge themselves to figure out the reasons before the author explains. People label this type of sci-fi as "hard sci-fi." + +Some say after drinking good pure tea, you can't stomach bubble tea anymore. My tea appreciation hasn't reached that level. But after finishing this book and thinking back to aliens I've seen before, I can't stand them anymore. Aliens won't suddenly appear over New York opening portals, America won't casually shoot two nuclear missiles then say "oh no, they have shields"—such depictions are too lazy. + +Where do humans and aliens meet? How do they communicate after meeting? Will it be like "The Three-Body Problem" where they must kill each other on sight? Or like "Arrival" where they gesticulate trying to teach aliens English? This book's answers are much more reasonable, with both IQ and EQ online. + +The novel was published in 2021, so probably written in 2019-2020. There was no ChatGPT explosion yet. The novel cleverly avoids too much computer technology, not needing to make empty capability estimates. + +The protagonist's superior gets enormous authorization from world leaders, able to ignore any procedures and oversight mechanisms to do anything desired. But faces accountability after the spaceship launches. Is this the author's ideal Earth response mechanism? The author doesn't detail these cooperation mechanisms much—the social science black box is very thick. Facing Earth's potential social impact, the superior recruits a bunch of natural scientists and climatologists but no historians, economists, or political experts. Social science students want to hold the author accountable a bit. + +The author places the protagonist's science team on China's aircraft carrier and Russia's base. When writing, the Russia-Ukraine war hadn't started yet—I wonder if the author thinks cooperating with authoritarian alliances is acceptable or inevitable when Earth faces major crisis? Will Hollywood movies follow this setting? diff --git a/content/posts/2025-11-21_project_hail_mary.zh-TW.md b/content/posts/2025-11-21_project_hail_mary.zh-TW.md new file mode 100644 index 0000000..5b8a07e --- /dev/null +++ b/content/posts/2025-11-21_project_hail_mary.zh-TW.md @@ -0,0 +1,35 @@ +--- +title: "《極限返航》閱讀心得" +date: 2025-11-21 +draft: false +--- + +![](https://books.google.com/books/content?id=-Ff2DwAAQBAJ&printsec=frontcover&img=1&zoom=1&edge=none&source=gbs_api) + +Project Hail Mary + +電影翻「極限返航」，2026 會上映 + +我本來想說我不會再看什麼太空人和外星人的科幻小說了，不過因為是禮物，我就讀了，也很好讀。 + +書的內容超容易爆雷，但電影預告片已經爆了不少，本文就以下內容和不影響劇情進展的部分談論。 + +- 人類遇到外星人 +- 人類需要在外太空解決問題 +- 人類需要頃地球之力完成一個計劃 + +乍看之下，這是一個早已寫爛的題目，古今已經多少人類遇到外星人的小說和電影？為什麼 2020 要多看一本？看完之後是真的會覺得 Andy Weir 真的傾全地球之力拯救了這個類別。 + +這本不會是一本會想用「充滿想像力」形容的書，作者「想了很多」才是讀者感嘆的地方。書裡面引入的新科幻概念不多，但每個概念扎實的互相影響，讓作者塑造的科幻世界充滿實感。 + +科幻的黑箱薄到比一層細胞膜還薄。在小說中，主角面對的種種挑戰，都像是一道理工題目。書裡面看似任意丟出來的數字：長度、質量、時間、溫度等等，讀者代入公式都算得出來（如果沒提到主角需要用試算表算很久的話）。讀者也可以挑戰在作者解說前算出原因。人們給這種類型的科幻小說一個「硬科幻」這個標籤。 + +有人說喝過好喝的純茶，外面的手搖飲就會喝不下去。我是品茶沒到那個境界。但看完這本再去回想以前看過的外星人，就會覺得看不下去了。外星人不會突然出現在紐約上空開傳送門，美國不會隨便射兩顆核彈然後說「哎呀，他們有防護罩」，這種描寫太懶惰了。 + +人類和外星人在什麼地方見面？見面之後該怎麼交流？會像三體那樣見面就得互殘？或是異星入境那樣比手畫腳硬是要教會外星人英文？本書給的答案合理很多，雙商都有在線。 + +小說在 2021 發表，所以可能是 2019~2020 寫就。那時候還沒有 ChatGPT 大爆發。小說非常精明地避開太多電腦科技，不用做一些虛浮的能力推估。 + +主角的長官得到全世界首領極大的授權，可以無視任何流程和監督機制，做任何想做的事情。但在太空船出發之後要面對咎責。這是作者理想中地球的反應機制嗎？作者沒有太多對這些合作機制細節的交代，社會科學的黑箱非常厚。面對地球潛在的社會衝擊，長官招募了一堆自然科學專家、氣候學家，就都沒招歷史、經濟、政治的專家。學習社會科學的讀者想對作者咎責一下。 + +作者安排主角的科學團隊在中國的航母和俄羅斯的基地。寫作時候俄烏戰爭也還沒開打，不知道作者是否認為在地球面臨重大危機的時候，和獨裁者聯盟合作是一件可以接受的事或一種必然？好萊烏電影會照著這個設定演嗎？