7/24/2026
6/24/2026
AI Alignment
- 李珊珊,「從中學就開始培養 AI 天才?」前美國奧數主教練羅博深:這讓我脊背發涼,知識分子,2026.6.23 (Prof. Po-Shen Loh) (new)
知識分子:您的意思是,在高中階段就對 AI 人才進行針對性的超前培養,太早了?
羅博深:不只是太早的問題,而是這種「純技術導向」的教育結構本身就存在致命缺陷。
如果一個孩子從高中起就進入純理科的軌道,每天只跟演算法、模型和程式碼打交道,從來沒有讀過小說,從來沒有在具體的文學語境裡理解人類複雜的現實,不懂道德約束與同理心,那麼當他掌握了最尖端的技術力量,他可能會在無意間對整個社會造成無法挽回的傷害。
一個人如果擁有龐大的科技力量,卻對它的邊界和人文影響完全不了解,他根本不應該去接觸這些核心技術。這就像核武器一樣。
所以,每當我看到一個因為過早分軌、在知識結構裡完全缺失人文的理工科學生,我都會擔憂:他真的不應該繼續做前沿科技研究了。我們已經走到了一個技術極有可能危及全人類的節點,如果一個人的教育結構是這樣一條腿長、一條腿短的狀態,他就不應該被允許繼續往這條路上走。
知識分子:您曾提到,面對來自 AI 的挑戰,孩子們需要更加「Thoughtful」,能具體解釋一下嗎?羅博深:我習慣用英文原詞來定義它,翻譯成「深思熟慮」之類的,並不太準確。Thoughtful 的人有兩個核心特質。第一,有共情能力,體貼周到,真心希望讓別人快樂——不只是父母和朋友,也包括陌生人,這才是真正的利他精神。第二,肯真正地去思考——他們喜歡「理解」事物,而不只是「會做」。有些學生考試能拿高分,但只是機械地照著方法做,並不理解背後的邏輯。我要找的是那種喜歡用自己的獨立思維想清楚一件事、不滿足於別人給定的標準答案的人。對人和對問題都有洞察,只要具備這兩點,一個人在未來會成長得很好。
- Medha Bankhwal, Ankit Bisht, Michael Chui, Roger Roberts, and Ashley van Heteren, AI for social good: Improving lives and protecting the planet, McKinsey, 2024.
- Yoshua Bengio et al., Managing extreme AI risks amid rapid progress, Science, 20 May 2024, Vol 384, Issue 6698, pp. 842-845 (arXiv)
- Jiaming Ji et al., AI Alignment: A Comprehensive Survey, arXiv:2310.19852 (website)
- Jack Stilgoe, Technological risks are not the end of the world, Science, 18 Apr 2024, Vol 384, Issue 6693.
6/06/2026
Use LLM to learn *** and its impact
Research and development
- Marina Favaro and Jack Clark, When AI builds itself, Anthropic (new)
- As of May 2026, more than 80% of the code we merge into Anthropic’s codebase was authored by Claude.
- In the second quarter of 2026, the typical engineer was merging 8× as much code per day as they were in 2024.
- On the most open-ended tasks, Claude’s success rate reached 76% in May 2026, up 50 percentage points in six months.
- In this world, the pace of progress in AI development becomes determined entirely by the availability of compute (or the speed of discovering various efficiencies in algorithmic training or inference) for AI systems. Humans play a substantially diminished role in their development, likely moving most of our effort towards oversight, validation, and verification of an expanding “virtual lab” run by AI systems. We expect that systems capable of automated AI research and development would have skills that would transfer to the rest of science, allowing them to begin to revolutionize other fields.
- Dimitris Bertsimas and Georgios Margaritis, Robust and Adaptive Optimization under a Large Language Model Lens, arXiv:2501.00568.
5/11/2026
Summarized AI information from Stanford HAI and IEEE Spectrum
Want to understand the global AI landscape? These reports are fantastic for staying updated on state-of-the-art trends. I highly recommend them to all my students, especially for doing research and job interview prep.
Stanford HAI, The 2026 AI Index Report
Eliza Strickland, 12 Graphs That Explain the State of AI in 2025 Stanford’s AI Index tracks performance, investment, public opinion, and more, IEEE Spectrum, 07 Apr 2025
Eliza Strickland, The Top 10 AI Stories of 2024, IEEE Spectrum, 31 Dec 2024.
5/07/2026
捐款給大學
美國大學 (和其研究、人才培育、衍生公司) 持續進步的重要原因
Michael T. Nietzel, University Of Wisconsin Receives $100 Million For Its New AI College, Forbes, Apr 28, 2026 (一億美金)
4/01/2026
2/28/2026
我們的歲月靜好,是因為有人負重前行
2/21/2026
The coming wave
Mustafa Suleyman and Michael Bhaskar, The Coming Wave: Technology, Power, and the Twenty-first Century's Greatest Dilemma, Crown, September 5, 2023
洪慧芳譯,控制邊緣:未來科技與全球秩序的抉擇,感電出版,2024
11/12/2025
10/10/2025
10/03/2025
8/05/2025
7/24/2025
歲月靜好
參加了數百次的系務會議,針對各種大大小小的提案和投票,每位老師都會拿到紙本說明,由主持人或系助理先朗讀條文內容,讓每位老師表達不同的看法和意見,再投票。有時候,遇到爭論不休的條文,一時之間老師們找不到共識,也不會強行表決;會議主席會裁決,再找時間討論。這就是民主的程序。
也曾擔任兩年的校預算委員會委員,每學期開會十多次,都會邀請提案的單位前來說明。因為年度預算有限,頂多是刪減預算,從來沒有將預算刪成負數。
7/22/2025
7/21/2025
7/11/2025
The Batch by A. Ng
- )
- Large scale system: The system aggregates data generated by 240 million customers and 2 million store personnel, feeding applications that streamline operations among 100,000 suppliers, 150 distributors, and 10,000 retail venues in 19 countries.
6/27/2025
6/15/2025
世界經濟論壇 2025 年未來就業報告
World Economic Forum, The Future of Jobs Report 2025, 7 January 2025.
Technological change, geoeconomic fragmentation, economic uncertainty, demographic shifts and the green transition – individually and in combination are among the major drivers expected to shape and transform the global labour market by 2030. The Future of Jobs Report 2025 brings together the perspective of over 1,000 leading global employers—collectively representing more than 14 million workers across 22 industry clusters and 55 economies from around the world—to examine how these macrotrends impact jobs and skills, and the workforce transformation strategies employers plan to embark on in response, across the 2025 to 2030 timeframe.
