Friday, June 27, 2025

AI 半導體產業應用

深度研究: AI 正在改變半導體製造業

 據業界推估,台積電和三星等領先企業已成功應用遷移學習,將CVD設備的模型遷移至PVD設備,將檢測效率提高50%以上,同時減少80%的新模型訓練時間。



TSMC & ASML

晶圓檢測

實驗備忘錄





Wednesday, June 11, 2025

AI 對話撞到長度限制 max length reached

 每個對話( chat) 都有容許的記憶長度(context window),這跟人類是不一樣的,人類可以在老年時依然記得初戀,或是回想起小時候第一次抱起乳貓小三花時鮮明的情感的湧現,但AI的記憶會受到這個context window 的限制,比較誇張的比喻是金魚腦,正如金魚會忘記兩秒以前發生的事情,只活在當下,忘記過去。當你跟AI的對話長度接近這個記憶長度時,如果系統放任不管,AI 接下來會開始遺忘先前討論的事情,像一個失智者開始胡言亂語,明明剛剛在聊去年住在巴黎的事,突然改口說我這輩子沒有出過國,當然系統會去阻止這種事情發生。至此,你有三個選項:


1. 縮短您的訊息:請求更簡潔的回應或以較少的回應請求資訊。

2. 開始新的對話:開始一個新的聊天以重置(reset 歸零)長度限制。

3. 升級至Claude Pro:這個付費選項可以提供更高的訊息長度限制。(ChatGPT 也有類似 Pro 付費服務)

如需立即解決,我建議開啟一個新的對話,並以較少的回應請求內容。例如,可以一次請求報告的一或兩個部分,而不是整個文件。


無法繼續輸入,是context window 問題,與限流無關,但可能剛好某一個chat 的 context window 用完時你也剛好被限流

只要開啟新的對話,記憶歸零

Aesthetics

 







Tuesday, June 10, 2025

AI 的關鍵提問 Q&A

 規模的迷思 : AI 的規模與問題解決策略







1. AI 有分左腦跟右腦嗎? (in English)
AI 模型並沒有像人類那樣的左右腦區分。它們是為特定任務設計的人工神經網絡。雖然有些模型在某些任務上表現較好,但這並非源於像人類大腦那樣的左右腦分工。然而有些模型(4o)內容見長,有些(o1, o3-mini)以推理見長,前者類似右腦,後者類似左腦

2. 為什麼 ChatGPT 4(非 RL LLM)不擅長規劃?
我需要說明我無法確認其他 AI 模型的具體能力或進行直接比較。不過,一般來說,規劃是一項複雜的認知任務,需要幾個關鍵能力:
- 理解因果關係和時間關係
- 維持連貫的長期目標
- 適應不斷變化的環境
- 管理資源和限制

純粹基於下一個詞預測訓練的語言模型可能在這些方面存在困難,因為它們本身並沒有順序決策或維持長期連貫性的機制。

3. 規模的迷思:AI 的規模與問題解決策略
規模本身並不能保證更好的問題解決能力。雖然較大的模型通常在許多任務中表現更好,但架構、訓練方法和具體的優化策略同樣重要。目前有許多研究在探討模型規模與能力之間的關係,表明架構創新和訓練方法可能比純粹的規模更為關鍵。

4. 有了 AI 之後,大學的價值是什麼
以 MIT 研究材料科學為例:
- 大學教育提供:
  - 基礎理論知識
  - 實驗操作經驗
  - 研究方法訓練
  - 專業人脈網絡
  - 批判性思維能力
  - 團隊合作經驗

學習 AI 和攻讀碩士學位服務於不同目的。碩士學位提供專業知識和研究經驗,而 AI 知識提供分析和自動化工具。選擇取決於職業目標 - 這兩者並不互斥,反而可以相輔相成。

5. Claude 3.5 Sonnet 會推理嗎?
作為 Claude 3.5 Sonnet,我可以描述我的能力:我能進行邏輯推理、問題解決和分析。但我是通過模式識別和學習到的關聯來完成這些任務,而不是通過與人類相同的認知過程。

6. 關於 DeepSeek 的事實與非事實

What AI can do for students like you

 在未來AI 世代,在張老師的課程中,我們會一起來學:


使用AI 協助閱讀英文文獻


如何使用AI 協助撰寫文獻整理報告*


如何使用AI 協助撰寫小論文


如何使用AI 繪製工程圖,為你的程式,論文,報告加分


如何使用AI 尋找研究新點子 (demo)


如何使用 AI 協助開發並且優化軟體程式引言)*



最好的程式語言





 寫程式還是嘴程式


早上花了一個多小時學習 Claude code, 出一張嘴就可以寫程式,debug, 測試,然後部署到平台。

雖然之前用過Claude code 很厲害,但上過這個短課程有了更深刻的理解。

最神奇的是,產生了洋洋灑灑的程式碼,卻沒有寫一行程式,以後寫程式可能要叫做「嘴程式」了。


Claude code 免費線上課程


AI很厲害,我們學過的他都會做,他目前唯一缺少的是意圖

例如他不會主動說來聊天吧,他也不會突然想寫一支無人機偵測程式

當然有一天也許他會有意識,至少目前沒有


所以現在做研究就是看你想得到什麼,就做得出什麼

不像以前你可能想做也做不出來

所以老師跟我的學生們與其說做研究,其實玩得很開心,不像苦命研究生


今日的世界,做苦工不一定有收穫,懂的跟AI溝通,善用AI,把AI當作我們的翅膀,任意翱翔天空


多個限制條件的複雜問題

  當AI面對包含多個限制條件的複雜問題時,確實存在難度。

在注意力機制(Attention mechanism)中,模型需要同時關注多個條件,並且每個注意力分配都是基於概率的。當條件數量增加時,要同時滿足所有條件的概率會顯著降低,這就像是連續投擲硬幣並期望全部正面朝上一樣——條件越多,全部滿足的概率越低。

這種情況下,模型可能會:

  • 忽略某些條件
  • 錯誤理解條件之間的關係
  • 在滿足一些條件的同時違反其他條件

如果您有一個包含多個複雜條件的問題需要解決,我建議將問題分解成較小的部分,逐步處理每個條件,這樣會更容易得到準確的解答。




一輛最多可載3人的汽車需要運送5個人 - Alice、Bob、Carlos、Diana 和 Elena。請考慮以下條件:

  • Alice 拒絕和 Bob 同車
  • Carlos 必須與 Diana 同行
  • Elena 是唯一持有駕照的人
  • 在起點或終點都不能留下單獨一位女性

    請擬定一個運輸計畫,能夠減少運輸人次更好。 
    failed plan normal

    Neutron star collision

      (story from Nature)










    Big data

     Grok to DeepSearch 66 sources






    Scientific Visualization

       1. 科學現象3d模擬(雷射,超導體,超流體,中子星碰撞...),選取以下一個題目,加上自選一個主題


    a. 科學模擬 3d 化(使用 three.js 模擬 3d )

    Black hole-star binary system simulation 

    黑洞與恆星雙星系統




    b.  使用 three.js 模擬 3d 


    2. 雷射(2d 模擬 energy pumping)




    雷射(3d 模擬) 使用3d量子場









    CO2 Sequestration Process, in 3D simulation.  二氧化碳封存地底岩層過程

    Strategic thinking

     

      Devise a strategy for the dog in the picture to retrieve its ball that fell into the pond without getting its body wet, with no humans nearby to help."

    This is a problem-solving or creative thinking exercise asking someone to come up with a plan for how a dog could get its ball back from water without getting wet and without human assistance.

    Animate your strategy in svg or in 3d.





    share (unsatisfactory)

    Use of float(人機協作)

    Monday, June 09, 2025

    場域應用與服務設計 課程介紹

    開課日期:2025年9月9日 (每周二 晚間 6:55 開始三節課)

    課程目標:本課程旨在探討人工智慧(AI)的理論基礎、實際應用及對社會的影響。學生將通過互動式講座和實作練習,深入了解AI演算法、可視化技術、職場應用及社會創新項目。課程結束後,學生將能夠在專業、學術及日常情境中有效運用AI工具。

    為什麼要修這門課?

    • 相關性:獲得塑造產業與社會的AI趨勢、演算法及應用的尖端知識。
    • 實務技能:培養使用AI工具的實作經驗,包括提示工程、存取擴增生成(RAG)和思維鏈(CoT)推理。
    • 跨學科方法:探索AI在職場生產力、學術寫作及社會創新中的角色。
    • 面向未來:通過實務與理論洞察,為日益由AI技術驅動的世界做好準備。

    課程結構

    • 互動式格式:每週課程結合講座、討論和實作練習,促進主動學習。
    • 練習:實作作業以鞏固概念,包括AI增強寫作、可視化及職場模擬。
    • 評量:每週練習及期中休息(未指定考試,但有一次暫停課程)。
    • 補充講座:額外的主題包括商業策略、社會創新及API使用,以豐富課程內容。



    Weekly Schedule

    Week 1: September 9, 2025

    Topic: Course Introduction

    • Why AI matters: Opportunities and challenges
    • Overview of course objectives and structure
      Exercise: Whac-A-Mole (打地鼠) – Introduction to interactive problem-solving with AI
      Supplemental Lecture: Service Overview: 7 Powers of Business Strategies
    • Understanding why certain AI-driven services excel using strategic frameworks

    Week 2: September 16, 2025

    Topic: AI Development Trends

    • Overview of current AI advancements
    • Algorithms: Foundations and applications
    • Engineering simulations with AI
      Exercise: Visualization – Create a basic AI-driven visualization (e.g., data or simulation output)

    Week 3: September 23, 2025

    Topic: Effective Prompt Engineering

    • Crafting high-quality prompts for AI models
    • Introduction to Retrieval Augmented Generation (RAG) – Part 1: Theory
    • Visualization and animation of fractals
    • Engineering animations with AI
      Exercise: Enhance AI – Practice writing optimized prompts for AI models

    Week 4: September 30, 2025

    Topic: Retrieval Augmented Generation (RAG) – Part 2: Applications

    • Practical applications of RAG in real-world scenarios
      Exercise: Enhance AI – Apply RAG techniques to a given dataset or problem

    Week 5: October 7, 2025

    Topic: Chain of Thought (CoT) Reasoning

    • Understanding and applying CoT for complex problem-solving
      Exercise: Enhance AI – Use CoT to solve a multi-step reasoning task

    Week 6: October 14, 2025

    Topic: LLM-Assisted Academic Writing (Chain of Thought)

    • Leveraging Large Language Models (LLMs) for research and writing
      Exercise: Enhance Writing by AI – Draft a section of a paper using LLM assistance

    Week 7: October 21, 2025

    Topic: Advanced LLM Applications in Academic Work

    • Advanced techniques for using LLMs in research and publication
      Exercise: Advanced Writing by AI – Refine and expand an academic paper using LLMs

    Week 8: October 28, 2025

    Topic: Advantages and Limitations of LLMs

    • Strengths and weaknesses of LLMs in various contexts
      Exercise: Planning and Reasoning – Design a project plan using AI reasoning tools

    Week 9: November 4, 2025

    Midterm Break: No class (Pause for reflection or catch-up)

    Week 10: November 11, 2025

    Topic: AI in the Workplace (Part 1)

    • Practical applications of AI in professional settings
    • Advanced usage of Claude (Anthropic’s AI model)
      Exercise: SWOT Analysis and War-Gaming – Conduct a business strategy exercise using AI tools
      Supplemental Lecture: Using AI for Social Service Projects, Social Innovation, and Social Enterprises
    • Applying AI to drive societal impact and innovation

    Week 11: November 18, 2025

    Topic: AI in the Workplace (Part 2)

    • Continued exploration of AI tools for workplace efficiency
      Exercise: AI Workplace – Implement an AI solution for a workplace scenario

    Week 12: November 25, 2025

    Topic: AI in the Workplace (Part 3)

    • A* algorithm and heuristics in AI
    • River-Crossing Problem: Applying 2 heuristics
      Exercise: River-Crossing Problem 3 – Solve an advanced version of the problem using AI heuristics

    Week 13: December 2, 2025

    Topic: AI in the Workplace (Part 4)

    • Case studies and advanced workplace applications
      Exercise: AI Workplace – Design a workplace AI tool or process
      Supplemental Lecture: ChatGPT API for Python Workshop
    • Hands-on workshop for building AI applications using the ChatGPT API

    Week 14: December 9, 2025

    Topic: AI in Everyday Life

    • Practical uses of AI in daily routines (e.g., productivity, decision-making)
      Exercise: Everyday AI – Create a personal productivity tool using AI
      Supplemental Lecture: Using AI for Social Innovation
    • Further exploration of AI-driven social impact projects

    Week 15: December 16, 2025

    Topic: Open Session / Project Work

    • Students work on final projects or explore advanced AI topics
      Exercise: Project development or AI application of choice

    Week 16: December 23, 2025

    Topic: Project Presentations

    • Students present their AI projects or case studies
      Exercise: Finalize and present projects

    Week 17: December 30, 2025

    Topic: Course Wrap-Up

    • Review of key concepts and takeaways
    • Discussion on the future of AI in various domains
      Exercise: Reflection – Write a short essay on AI’s impact on your field of interest



    評分與期望

    • 參與度:20%(積極參與互動課程)
    • 每週練習:80%(完成實作作業)
    • 期望:參加所有課程,按時完成練習,並積極參與討論。具備AI概念或程式設計基礎知識有助於學習,但非必要。

    資源

    • 將提供或推薦使用AI工具(例如ChatGPT、Claude等)。
    • 課程期間將分享補充教材和程式碼範例。

    場域應用與服務設計 Syllabus 2025 Fall


    9/9

    什麼是「場域應用與服務設計」?AI 如何賦能服務設計? -> 介紹學期專案  

    AI 入門

    示範: AI 快速開發 無人機飛行模擬視覺化

    AI 在半導體產業的應用

    AI 在電子產業的應用

    百工如何使用AI(SVG)

    CL#1



    AI 發展趨勢 9/16

    AI 互動的基礎:如何與 AI 有效溝通 (Prompt 基礎篇)

    練習: 使用 Prompt 優化 AI 的回覆


    科學動畫 9/16

    CL#2


    9/23


    10/7

    讓 AI 擁有領域知識 -> RAG 理論與應用(Retrieval Augmented Generation) 

    練習: Understanding academic papers


    10/14, 10/21



    10/28

    Vibe Coding 與 AI 輔助開發: 軟體工程的革命

    Aesthetics




    期中考暫停上課一次 11/4


    11/11


    策略性思維(取物策略)

    練習: 規劃與推理 Help retrieve


     
    Mini Projects 12/2, 12/9





    學期專案發表與展示 (Final Project Presentations) 12/23


    學期專案發表與展示 (Final Project Presentations) 12/30