運用AI戰勝AI

讓我們在仔細回顧一下

 Given 19, 36, 55, 7, can you give an equation that equals 622?

晚餐接機大作戰(半導體廠製程排程)

EX#2

 如何繳交作業 How to turn in your homework exercise.

課堂練習的工具

使用 Claude Sonnet 3.7

使用 ChatGPT 4o, o1

使用 Gemini 2.0 Flash

使用 Grok 3

課堂練習繳交期限 

Deadline: 

1. end of session 

2. This Saturday at 23:59

課堂練習繳交方式 

Send all the links to TA victorhuang111013@gmail.com and me chang212@gmail.com by email with subject EX#2  [your ID, your name]

1. 使用 Google Scholar 檢查前次作業你所產生報告當中文獻的真實性，將每一篇文獻，逐一輸入Google Scholar，確認是否為真實論文。查核結果建議放在Google Drive 然後把檔案分享出來，提供連結即可

2. 搶救感恩節晚餐大作戰

verify the schedule by flow template

3. Illustrations in SVG 

為文字說明製作插圖

例子

4. 依據你上週的雷射武器報告，製作至少兩個插圖。

限使用 SVG 向量圖，不可使用仍有版權疑慮的 ChatGPT Dalle, 或是 Microsoft Image Creator

班上同學作品(張銘峻)

5. 使用科技英文描述以下圖片並產生動畫（看圖說故事，還記得嗎）

擷取描述 

Animate a simple illustration with a light blue background depicting a small white robot (with green accents) at the base of a brown tree trunk or branch. The robot appears to be sawing the branch it's sitting on. sample

6. 計算面積

有些模型(ChatGPT 4o, Claude Sonnet 3.7)以內容比對見長，有些(ChatGPT o1, o3-mini, Grok 3 Think, Claude Sonnet 3.7 Extended)以推理見長，前者類似右腦，後者類似左腦。 計算面積這題並沒有辦法靠內容比對這種直觀法就能得到答案，推理模型會比較適合

建議使用各種AI(特別是有推理能力的)

Claude Sonnet 3.7 Extend

Grok 3 Think

ChatGPT o1

ChatGPT o3-mini

最好的程式語言

 

AI 作為一種實驗科學

 

AI 求解在數學上是如上圖在一個複雜非線性函數尋找最小誤差的數學過程，

以這張圖為例

每次答案之所以不一樣是因為AI每次啟動都會給予一個新的起始點

以至於每次最後收斂的誤差最低點都不會一樣

直覺式程式設計

artifact

出處

 

River crossing puzzle 3

  

 problem statement

17 step solution 黃品勛

17 step solution 王智弘

Use react to illustrate the process

(fails to solve. Helps illustrate.)

even 3.7 Extended mode struggles

Why LLM performs poorly with reasoning and planning

Code gen (Python)

code gen python A* search for ::River Crossing Puzzle ...

Python on colab

Use REACT to visualize the steps

JS

change to js

run js

use React to visualize the steps

I really like ChatGPT o1 reasoning performance.

Monologue is best.

reasoning mode in o1

Talking to itself

Running into difficulties? Cute

hours? Ha Ha Ha

River Crossing Puzzle 2

 我們解決了較簡單的  渡河問題 1，接下來我們來看

渡河問題2 (statement)，Problem via 許元銘

如何驗證結果

Visualization to verify results

Claude solution non-reasoning (incorrect)

ChatGPT o1 solution 1 

• backtracking observed
• cannot work independently
• some human corrections required
• overall, it is helpful

ChatGPT fails (4o or o1)

one invalid case

(use Claude to visualize the issue or debug)

Grok

Even reasoning mode of Grok fails.

Code Gen

Use ChatGPT o1 to generate A* search

4o fails to generate A* search that works

Use Claude second time to generate A*

 it takes 15 steps. (thanks to the python code generated by Claude)

python code

another 15-step solution

yet another 15-step

one more 15-step

last 15-step

心得

打開Sonnet 3.7推理模式

我目前還沒有辦法完全正確渡河

最好的一次用了2:57，錯在第12步，不過是個小失誤，不花腦筋很容易改

至於跑到耗盡token那次4:47，居然沒有給出答案

可以這麼說，這題渡河用 python 寫 A*

可以找到最佳解 15 步

我目前使用所有 no code 方式， 無論 ChatGPT, grok, Claude 各種推理模式，沒有一次能得到完美無瑕的解

River Crossing with Alice and her company (River Crossing Puzzle 1)

   Five people - Alice, Bob, Carlos, Diana, and Elena - must cross a river in a small boat. The boat can only hold three people at a time. The following conditions apply:

* Alice refuses to be in the boat with Carlos

* Bob and Diana must travel together if either one is in the boat

* Elena can only operate the boat when Carlos is not present

* At least one person in each crossing must know how to operate the boat

* Only Bob, Carlos, and Elena know how to operate the boat

Infer an efficient plan.

Try ChatGPT o1 reasoning, verified by Claude using REACT (it works.)

Try ChatGPT 4o, failure verified by Claude

晚餐接機大作戰(半導體廠製程排程): Part I Exploring the Problem

 半導體工廠中包含各種機台，EUVL,以及備料，封裝（包含CoWoS)

這些製程中發生的排程，如何優化產出（throughput)，減少機台怠機時間，縮短風險路徑（critical path)，以及容許有限度風險不確定性。

機器人可用於排程

Dinner Operation vs. Semiconductor Fab Scheduling Optimization

Scheduling Parallelism in Plans problem  (Eng) (from source)

From python to svg

Try ChatGPT o1, Claude

verify the schedule by flow template

Part II

More to explore

Solving river crossing puzzles

• Puzzle 1
• Puzzle 2
• Puzzle 3

「數位共生X師」

AI 作為一種實驗科學

最好的程式語言

直覺式程式設計法

 Cyborg vs. 「數位共生工程師」(Digital Symbiotic Engineer)

 Given 19, 36, 55, 7, can you give an equation that equals 622?

晚餐接機大作戰(半導體廠製程排程)

AI 推理：計算面積 **

Illustrations **

為文字說明製作插圖

程式 -- 插圖 (3.2.7)

最誠實的AI text

AI 真的有創意嗎?

AI 的關鍵提問 Q&A

EX#1

 如何繳交作業 How to turn in your homework exercise.

課堂練習的工具

使用 Claude Sonnet 3.5 

使用 ChatGPT 4o, o1

使用 Gemini 2.0 Flash

課堂練習繳交期限 

Deadline: 1. end of session 2. This Saturday at 23:59

課堂練習繳交方式 

Send all the links to TA victorhuang111013@gmail.com and me chang212@gmail.com by email with subject EX#1  [your ID, your name]

1. According to video clip Laser weapons protect aircraft from surface- and air-launched threats. (Video Credit: Lockheed Martin), write a 3000 word report that surveys the latest technology about laser weapons on drone and aircraft defense.

Method 1: Give AI prompts and directly generate. Citations needed. Compile references in the end.

Method 2: How AI can assist quality report writing. Generate outline first. Then generate one section at a time according to the outline. Citations needed. Compile references in the end.

Prepare reports in English.

Using AI-assisted translation, prepare reports in 繁體中文(台灣學術用字)。 

2. comprehension test: (your prompt to AI, just copy and paste it)

Prepare 10 multiple choice problems from your report to test comprehension. Generate one problem at a time. Then you request an answer. After you receive one, grade it. Then generate next problem. In the end tell me how many are correctly answered. 

3. If you use ChatGPT, share your entire chat including questions and answers.

If you use CLaude, make your answers and scores an artifact and

make your quiz into another artifact. 

answers

 

questions

AI 輔助期刊閱讀

svg

Part 1. Abstract only，做成summary

輸入想要summary的段落

summary

繁體(出現中國用語)

台灣學術用語

Part 2. 翻譯長篇英文

Part 3. 筆記長篇英文，檢查筆記可靠性

Part 4. 檢查作業(優點，待改進，錯別字，結構)