#tw-popsci
2023-11-02
我讀了這篇：*微軟或成 OpenAI 投資案的最大贏家*
讀完之後我再請chatGPT整理
我的prompt:
"""分析這篇文章
列出
1) 哪些是本篇作者獨創的想法
2)哪些是整理蒐集到資訊的轉述和整理，並附上reference
3)後續有什麼值得持續觀察的點，請列出時間點，和理由
4)分析本篇文章的結構，易讀性，給予一些改善的建議"""
CHATGPT的結果
1. 哪些是本篇作者獨創的想法:
◦ 微軟在 Satya Nadella 擔任 CEO 期間的投資與策略合作評價。
◦ 10B 美金相對於其他大型投資或項目的價值（如台積電的晶圓廠或元宇宙）的比較。
◦ 對微軟是否划算的質疑與推測。
◦ 結合多個資訊源來評估微軟的投資回報。
2. 哪些是整理蒐集到資訊的轉述和整理，並附上reference:
◦ OpenAI 在 2019 年的拆分信息。
◦ 微軟與 OpenAI 的合作細節。
◦ Source: Fortune - 關於微軟和 OpenAI 的投資結構。
◦ 路透社的消息指出，OpenAI 2023 和 2024 的目標營收。
◦ 2019 年的公開信提及 FCP 可以拿回的報酬。
◦ Pitchbook 上的數字關於 FCP 的分潤。
3. 後續有什麼值得持續觀察的點，請列出時間點，和理由:
◦ 2023/2024: OpenAI 的目標營收。理由：這將決定微軟的投資回報。
◦ 未來 4~7 年: 觀察微軟的投資回本情況。理由：評估投資的成功率。
◦ 未來的任何公開信或報告：更新 FCP 和微軟的分潤情況。理由：這將有助於確定微軟的實際獲利。
4. 分析本篇文章的結構，易讀性，給予一些改善的建議:
◦ 結構: 文章首先討論微軟的投資策略，然後深入探討與 OpenAI 的合作和投資回報。
◦ 易讀性: 適中。文章使用了清晰的段落和標題來組織信息，但部分數據和細節可能會讓非專業的讀者感到困惑。
▪︎ 建議:加入一些視覺元素，例如圖表或圖形，以幫助讀者更好地理解投資和回報的關係。
▪︎ 簡化部分專業術語或提供解釋，使其更易於大眾理解。
◦ 考慮增加一個總結部分，重點強調文章的主要觀點和結論
我個人覺得這篇給我很多新的訊息，我最喜歡你選的Fortune 雜誌的那張圖，非常有趣，高資訊量．
但我覺得medium的排版有點花．不知道是否是我不習慣而已．
讀完之後我再請chatGPT整理
我的prompt:
"""分析這篇文章
列出
1) 哪些是本篇作者獨創的想法
2)哪些是整理蒐集到資訊的轉述和整理，並附上reference
3)後續有什麼值得持續觀察的點，請列出時間點，和理由
4)分析本篇文章的結構，易讀性，給予一些改善的建議"""
CHATGPT的結果
1. 哪些是本篇作者獨創的想法:
◦ 微軟在 Satya Nadella 擔任 CEO 期間的投資與策略合作評價。
◦ 10B 美金相對於其他大型投資或項目的價值（如台積電的晶圓廠或元宇宙）的比較。
◦ 對微軟是否划算的質疑與推測。
◦ 結合多個資訊源來評估微軟的投資回報。
2. 哪些是整理蒐集到資訊的轉述和整理，並附上reference:
◦ OpenAI 在 2019 年的拆分信息。
◦ 微軟與 OpenAI 的合作細節。
◦ Source: Fortune - 關於微軟和 OpenAI 的投資結構。
◦ 路透社的消息指出，OpenAI 2023 和 2024 的目標營收。
◦ 2019 年的公開信提及 FCP 可以拿回的報酬。
◦ Pitchbook 上的數字關於 FCP 的分潤。
3. 後續有什麼值得持續觀察的點，請列出時間點，和理由:
◦ 2023/2024: OpenAI 的目標營收。理由：這將決定微軟的投資回報。
◦ 未來 4~7 年: 觀察微軟的投資回本情況。理由：評估投資的成功率。
◦ 未來的任何公開信或報告：更新 FCP 和微軟的分潤情況。理由：這將有助於確定微軟的實際獲利。
4. 分析本篇文章的結構，易讀性，給予一些改善的建議:
◦ 結構: 文章首先討論微軟的投資策略，然後深入探討與 OpenAI 的合作和投資回報。
◦ 易讀性: 適中。文章使用了清晰的段落和標題來組織信息，但部分數據和細節可能會讓非專業的讀者感到困惑。
▪︎ 建議:加入一些視覺元素，例如圖表或圖形，以幫助讀者更好地理解投資和回報的關係。
▪︎ 簡化部分專業術語或提供解釋，使其更易於大眾理解。
◦ 考慮增加一個總結部分，重點強調文章的主要觀點和結論
我個人覺得這篇給我很多新的訊息，我最喜歡你選的Fortune 雜誌的那張圖，非常有趣，高資訊量．
但我覺得medium的排版有點花．不知道是否是我不習慣而已．
我讀了Louis的一篇文章，想嘗試說明我想做的事情：
1. 大量sourcing
2. 自動，產生系統性思考prompt，整理，挑選，這邊的prompt理想中應該是會持續被使用，也可以共同一起review
3. 根據自動產生的整理，最後加上一點點我自己針對這篇文章的看法，角度和感覺(commets)．
1. 大量sourcing
2. 自動，產生系統性思考prompt，整理，挑選，這邊的prompt理想中應該是會持續被使用，也可以共同一起review
3. 根據自動產生的整理，最後加上一點點我自己針對這篇文章的看法，角度和感覺(commets)．
- 👍1
我前面分享的天下雜誌 app 也嘗試在做這樣的事，我初用一天的感覺是： AI 還是廢話太多。
有沒有可能，透過 prompt 之後，產生一些關鍵字或是關鍵句子？
甚至在 AI 閱讀多篇新聞之後 (跨度可能是好幾天/星期/月) ，把訊息之間的關係用圖表整理出來，目地是「提供脈絡」。
以 OpenAI + 微軟為例，整個脈絡應該要包括一開始的合作，到後來 ChatGPT 整碗被微軟端走，互相變成利益衝突關係。
有沒有可能，透過 prompt 之後，產生一些關鍵字或是關鍵句子？
甚至在 AI 閱讀多篇新聞之後 (跨度可能是好幾天/星期/月) ，把訊息之間的關係用圖表整理出來，目地是「提供脈絡」。
以 OpenAI + 微軟為例，整個脈絡應該要包括一開始的合作，到後來 ChatGPT 整碗被微軟端走，互相變成利益衝突關係。
AI擅長的事：大量&廣泛的sourcing，自動產生通順的內容
AI不擅長的事：如果本來就沒有好的source，也不會產生更好的內容
如果拿這一套系統用作對於非常大量source（例如說，整個基礎科學領域這週有什麼突破性研究？）的 daily/weekly/monthly 新知的整理+newsletter感覺會非常不錯，我們只需要做最後把關的editing。但是要叫AI把難的內容講簡單可能還是有點距離。
如果是做automated news in all fields感覺會很有趣而且也很valuable，也就是你提到的第三點-自動化整理加上我們的把關跟一點點comments
AI不擅長的事：如果本來就沒有好的source，也不會產生更好的內容
如果拿這一套系統用作對於非常大量source（例如說，整個基礎科學領域這週有什麼突破性研究？）的 daily/weekly/monthly 新知的整理+newsletter感覺會非常不錯，我們只需要做最後把關的editing。但是要叫AI把難的內容講簡單可能還是有點距離。
如果是做automated news in all fields感覺會很有趣而且也很valuable，也就是你提到的第三點-自動化整理加上我們的把關跟一點點comments
`要叫AI把難的內容講簡單可能還是有點距離。`
please explain generative AI in 5 levels:
5 years old pre-K student
8 years old 2rd grade
12 years old 8th grade
18 years old high school
24 years old college student
ChatGPT
Sure! Let's break it down:
For a 5-year-old pre-K student:
Imagine you have a magic coloring book. When you tell it, "I want a picture of a dragon," it listens and creates a new dragon picture for you. It can make many different pictures because it's learned from looking at lots and lots of other drawings.
For an 8-year-old 2nd grader:
You know how you can use your imagination to think of a story or a picture? Well, generative AI is like a computer that can use its imagination. You give it an idea, like "a robot playing soccer," and it uses all the things it has seen before to make a new picture or story about a soccer-playing robot.
For a 12-year-old 8th grader:
Generative AI is a type of computer program that can create new things - like pictures, music, or even stories - that it's never seen before. It works by learning from lots of examples, like how you learn in school. So, if you've been learning to draw by looking at many pictures, you can try to draw something new. The AI does that, but it uses data and patterns to make something unique.
For an 18-year-old high school student:
Generative AI refers to the artificial intelligence algorithms that can generate new content. It's based on machine learning, where the system is fed a large amount of data, like images or texts, and it finds patterns in that data. Then, it can use those patterns to create new things that it hasn't been explicitly programmed to make. It's like how you can learn the rules of grammar and use them to write sentences you've never heard before.
For a 24-year-old college student:
Generative AI encompasses advanced algorithms within machine learning that are designed to produce new content, whether it be in the form of text, images, or sounds. This is achieved through models like Generative Adversarial Networks (GANs) or Variational Autoencoders (VAEs), which essentially learn the underlying distribution of the input data. These models can then sample from this learned distribution to generate novel outputs that are statistically similar but not identical to the data they were trained on. The power of generative AI lies in its ability to interpolate and extrapolate from complex datasets, allowing for the creation of original and coherent outputs that can mimic human creativity.
5 years old pre-K student
8 years old 2rd grade
12 years old 8th grade
18 years old high school
24 years old college student
ChatGPT
Sure! Let's break it down:
For a 5-year-old pre-K student:
Imagine you have a magic coloring book. When you tell it, "I want a picture of a dragon," it listens and creates a new dragon picture for you. It can make many different pictures because it's learned from looking at lots and lots of other drawings.
For an 8-year-old 2nd grader:
You know how you can use your imagination to think of a story or a picture? Well, generative AI is like a computer that can use its imagination. You give it an idea, like "a robot playing soccer," and it uses all the things it has seen before to make a new picture or story about a soccer-playing robot.
For a 12-year-old 8th grader:
Generative AI is a type of computer program that can create new things - like pictures, music, or even stories - that it's never seen before. It works by learning from lots of examples, like how you learn in school. So, if you've been learning to draw by looking at many pictures, you can try to draw something new. The AI does that, but it uses data and patterns to make something unique.
For an 18-year-old high school student:
Generative AI refers to the artificial intelligence algorithms that can generate new content. It's based on machine learning, where the system is fed a large amount of data, like images or texts, and it finds patterns in that data. Then, it can use those patterns to create new things that it hasn't been explicitly programmed to make. It's like how you can learn the rules of grammar and use them to write sentences you've never heard before.
For a 24-year-old college student:
Generative AI encompasses advanced algorithms within machine learning that are designed to produce new content, whether it be in the form of text, images, or sounds. This is achieved through models like Generative Adversarial Networks (GANs) or Variational Autoencoders (VAEs), which essentially learn the underlying distribution of the input data. These models can then sample from this learned distribution to generate novel outputs that are statistically similar but not identical to the data they were trained on. The power of generative AI lies in its ability to interpolate and extrapolate from complex datasets, allowing for the creation of original and coherent outputs that can mimic human creativity.
我叫AI看research paper還是看的零零落落的 🤣 都給我換句話亂講哈哈哈
不過叫AI角色扮演/換句話說已經有體系的內容就會很棒
感謝你的回饋，非常有建設性 (也包括 chatgpt 的分析)。
follow-up 1: 如果你希望 AI 幫你統整，你覺得 1-4 那個問題比較重要？
follow-up 2: 你方便多描述「medium的排版有點花」這部份嗎？ (很歡迎主觀意見) UX 也是很重要的。
四個問題背後的目的
1) 哪些是本篇作者獨創的想法 -> 幫我我瞭解作者，主觀決定是否要訂閱他的頻道
2)哪些是整理蒐集到資訊的轉述和整理，並附上reference -> 幫我了解上游，了解資料的源頭，也許我以後會想去源頭看看有什麼有趣，但沒有被你選到的文章
3)後續有什麼值得持續觀察的點，請列出時間點，和理由-->判斷是否有action item
4)分析本篇文章的結構，易讀性，給予一些改善的建議 --> 想想我自己能寫得更好嗎
1) 哪些是本篇作者獨創的想法 -> 幫我我瞭解作者，主觀決定是否要訂閱他的頻道
2)哪些是整理蒐集到資訊的轉述和整理，並附上reference -> 幫我了解上游，了解資料的源頭，也許我以後會想去源頭看看有什麼有趣，但沒有被你選到的文章
3)後續有什麼值得持續觀察的點，請列出時間點，和理由-->判斷是否有action item
4)分析本篇文章的結構，易讀性，給予一些改善的建議 --> 想想我自己能寫得更好嗎
UX上我想了一想還是講不大出來，得請UX大神幫忙了
我前面分享的天下雜誌 app 也嘗試在做這樣的事，我初用一天的感覺是： AI 還是廢話太多。
有沒有可能，透過 prompt 之後，產生一些關鍵字或是關鍵句子？
甚至在 AI 閱讀多篇新聞之後 (跨度可能是好幾天/星期/月) ，把訊息之間的關係用圖表整理出來，目地是「提供脈絡」。
以 OpenAI + 微軟為例，整個脈絡應該要包括一開始的合作，到後來 ChatGPT 整碗被微軟端走，互相變成利益衝突關係。
有沒有可能，透過 prompt 之後，產生一些關鍵字或是關鍵句子？
甚至在 AI 閱讀多篇新聞之後 (跨度可能是好幾天/星期/月) ，把訊息之間的關係用圖表整理出來，目地是「提供脈絡」。
以 OpenAI + 微軟為例，整個脈絡應該要包括一開始的合作，到後來 ChatGPT 整碗被微軟端走，互相變成利益衝突關係。
那我換個方式問好了 XD 同樣一篇文章，媒體編輯潤飾過後，套用該網站範本，有沒有比較好閱讀？ https://www.bnext.com.tw/article/74124/deal-structure-2019-between-openai-early-investor
有 XD
我仔細看了一下，提供一些主觀感受和意見
1. 我覺得字體太大
2. 標題用斜體感覺不正式
3. 好幾張圖解析度不對
4. 我覺得bnext的title有加[tag]會更容易閱讀
我仔細看了一下，提供一些主觀感受和意見
1. 我覺得字體太大
2. 標題用斜體感覺不正式
3. 好幾張圖解析度不對
4. 我覺得bnext的title有加[tag]會更容易閱讀
啊，這應該是 Medium 手機版自己幫我加上去的，我沒瓣法控制。 桌機版沒有你說的問題~
2023-11-03
AI擅長的事：大量&廣泛的sourcing，自動產生通順的內容
AI不擅長的事：如果本來就沒有好的source，也不會產生更好的內容
如果拿這一套系統用作對於非常大量source（例如說，整個基礎科學領域這週有什麼突破性研究？）的 daily/weekly/monthly 新知的整理+newsletter感覺會非常不錯，我們只需要做最後把關的editing。但是要叫AI把難的內容講簡單可能還是有點距離。
如果是做automated news in all fields感覺會很有趣而且也很valuable，也就是你提到的第三點-自動化整理加上我們的把關跟一點點comments
AI不擅長的事：如果本來就沒有好的source，也不會產生更好的內容
如果拿這一套系統用作對於非常大量source（例如說，整個基礎科學領域這週有什麼突破性研究？）的 daily/weekly/monthly 新知的整理+newsletter感覺會非常不錯，我們只需要做最後把關的editing。但是要叫AI把難的內容講簡單可能還是有點距離。
如果是做automated news in all fields感覺會很有趣而且也很valuable，也就是你提到的第三點-自動化整理加上我們的把關跟一點點comments
@emily.kuowen has joined the channel
`要叫AI把難的內容講簡單可能還是有點距離。`
please explain generative AI in 5 levels:
5 years old pre-K student
8 years old 2rd grade
12 years old 8th grade
18 years old high school
24 years old college student
ChatGPT
Sure! Let's break it down:
For a 5-year-old pre-K student:
Imagine you have a magic coloring book. When you tell it, "I want a picture of a dragon," it listens and creates a new dragon picture for you. It can make many different pictures because it's learned from looking at lots and lots of other drawings.
For an 8-year-old 2nd grader:
You know how you can use your imagination to think of a story or a picture? Well, generative AI is like a computer that can use its imagination. You give it an idea, like "a robot playing soccer," and it uses all the things it has seen before to make a new picture or story about a soccer-playing robot.
For a 12-year-old 8th grader:
Generative AI is a type of computer program that can create new things - like pictures, music, or even stories - that it's never seen before. It works by learning from lots of examples, like how you learn in school. So, if you've been learning to draw by looking at many pictures, you can try to draw something new. The AI does that, but it uses data and patterns to make something unique.
For an 18-year-old high school student:
Generative AI refers to the artificial intelligence algorithms that can generate new content. It's based on machine learning, where the system is fed a large amount of data, like images or texts, and it finds patterns in that data. Then, it can use those patterns to create new things that it hasn't been explicitly programmed to make. It's like how you can learn the rules of grammar and use them to write sentences you've never heard before.
For a 24-year-old college student:
Generative AI encompasses advanced algorithms within machine learning that are designed to produce new content, whether it be in the form of text, images, or sounds. This is achieved through models like Generative Adversarial Networks (GANs) or Variational Autoencoders (VAEs), which essentially learn the underlying distribution of the input data. These models can then sample from this learned distribution to generate novel outputs that are statistically similar but not identical to the data they were trained on. The power of generative AI lies in its ability to interpolate and extrapolate from complex datasets, allowing for the creation of original and coherent outputs that can mimic human creativity.
5 years old pre-K student
8 years old 2rd grade
12 years old 8th grade
18 years old high school
24 years old college student
ChatGPT
Sure! Let's break it down:
For a 5-year-old pre-K student:
Imagine you have a magic coloring book. When you tell it, "I want a picture of a dragon," it listens and creates a new dragon picture for you. It can make many different pictures because it's learned from looking at lots and lots of other drawings.
For an 8-year-old 2nd grader:
You know how you can use your imagination to think of a story or a picture? Well, generative AI is like a computer that can use its imagination. You give it an idea, like "a robot playing soccer," and it uses all the things it has seen before to make a new picture or story about a soccer-playing robot.
For a 12-year-old 8th grader:
Generative AI is a type of computer program that can create new things - like pictures, music, or even stories - that it's never seen before. It works by learning from lots of examples, like how you learn in school. So, if you've been learning to draw by looking at many pictures, you can try to draw something new. The AI does that, but it uses data and patterns to make something unique.
For an 18-year-old high school student:
Generative AI refers to the artificial intelligence algorithms that can generate new content. It's based on machine learning, where the system is fed a large amount of data, like images or texts, and it finds patterns in that data. Then, it can use those patterns to create new things that it hasn't been explicitly programmed to make. It's like how you can learn the rules of grammar and use them to write sentences you've never heard before.
For a 24-year-old college student:
Generative AI encompasses advanced algorithms within machine learning that are designed to produce new content, whether it be in the form of text, images, or sounds. This is achieved through models like Generative Adversarial Networks (GANs) or Variational Autoencoders (VAEs), which essentially learn the underlying distribution of the input data. These models can then sample from this learned distribution to generate novel outputs that are statistically similar but not identical to the data they were trained on. The power of generative AI lies in its ability to interpolate and extrapolate from complex datasets, allowing for the creation of original and coherent outputs that can mimic human creativity.
我叫AI看research paper還是看的零零落落的 🤣 都給我換句話亂講哈哈哈
不過叫AI角色扮演/換句話說已經有體系的內容就會很棒
2023-11-06
@tiffany1998wang
這個【你我正在使用的 AI 熟讀大量的中國開放資料，再*簡轉繁*給我們使用...】
是妳的提案動機，觀察到現象的加速過程，
先丟在這個tread給坑主參考，相對偏題。
只是文章內的『台灣要訓練自己的 AI ，第一關是「大量的，好 幾百TB 的開源正體中文資料集」』
我覺得跟原提案的目標挺類似，而且方法挺系統性的。
這個【你我正在使用的 AI 熟讀大量的中國開放資料，再*簡轉繁*給我們使用...】
是妳的提案動機，觀察到現象的加速過程，
先丟在這個tread給坑主參考，相對偏題。
只是文章內的『台灣要訓練自己的 AI ，第一關是「大量的，好 幾百TB 的開源正體中文資料集」』
我覺得跟原提案的目標挺類似，而且方法挺系統性的。
wow非常類似，但仔細讀下去他也沒有提到太具體可行的執行方式，應該也是需要很多人力一起合作，不過我們可以畫一個執行流程圖方便大家一起參與進來， 我想一下
2023-11-08
any AI gadgets on openai store that’s potentially useful? XD
2023-11-09
any AI gadgets on openai store that’s potentially useful? XD