AI逐字稿_黃仁勳推AI民主化背後挑戰：如何不讓數位落差擴大？
AI Transcript: Jensen Huang on the Challenges of AI Democratization: How to Prevent the Widening of the Digital Divide

人工智能(AI)讓我變得更聰明，原因是我現在擁有多位導師(tutors)。我每天都使用人工智能作為學習的輔導工具。我使用了三種人工智能工具：我經常使用ChatGPT，還會使用Gemini Pro和Perplexity。這三種人工智能工具都非常出色。

AI has made me smarter, and the reason for that is because I now have tutors. I use my AIs for tutors for tutoring every day. I have three AI tools: I use ChatGPT a lot, I use Gemini Pro, and I use Perplexity. And these three AIs are excellent.

最近，我也開始使用Grok。這些人工智能(AI)都非常出色，每一種都有其略微不同的風格。因此，你需要了解它們各自的風格。請注意，我使用的不是一種，而是四種人工智能工具。原因在於它們各有不同的技能和風格。根據我想完成的任務，我會選擇合適的人工智能工具，就像根據我需要完成的工作，我會去找我的管理團隊(management team)中最適合的成員一樣。對，就是這樣。

And recently, I've been using Grok as well. These are excellent AIs, and each one of them has its slightly different style. And so, you just have to learn their style. Notice I have four AIs, not one. And the reason for that is they all have different skills and different styles. And depending on what I want to do, I go to the right person, just like depending on what I need to have done, I go to the right member of my management team. yeah. yeah. Same thing.

好的，在接下來的部分，我主動聯繫了兩位來自台灣的本地學生，他們來自兩個截然不同的領域：一位主修人文学科(humanities)，另一位主修電腦工程(computer engineering)。我詢問了他們對你的問題。你有興趣聽聽他們的問題嗎？第一位學生名叫詹勳澔(Jen Shinhao)。他是國立台灣大學電腦科學與資訊工程研究所(Graduate School of Computer Science and Information Engineering)的一年級碩士生。他的問題有些長，但我會完整讀出。他提到，你曾談到人工智能(AI)的民主化是一股變革力量，能夠在全球範圍內傳播智慧、機會和能力。然而，一些觀察家擔憂，這類強大的技術可能會擴大數位鴻溝(digital divide)，而不是縮小它。從你的角度來看，應該採取哪些措施來確保這場人工智能革命惠及廣大群眾而非少數人？我們又必須共同克服哪些挑戰來實現這一願景？

Okay. So for the next segment, I've actually taken the liberty of contacting two local students in Taiwan from two very different spectrum: one studying humanities and one studying computer engineering. I asked them what questions they have for you. We want to be interested to listen to what they have. Okay So, for the first one, his name is Jen Shinhao Shinhao Chang(詹勳澔). He’s a first-year master’s student at the Graduate School of Computer Science and Information Engineering at National Taiwan University. He has a pretty long question, but I’ll just read it out. He said that you have spoken about the democratization of AI as a transformative force, capable of spreading intelligence, opportunity, and capability around the world. Yet, some observers are concerned that such technologies, while powerful, might exacerbate the digital divide rather than narrow it. So, from your perspective, what must be done to ensure that this AI revolution benefits the many rather than the few, and what challenges must be collectively overcome to make this vision a reality?

人工智能(AI)的民主化極為重要，這是無庸置疑的。這就是為什麼每個地區、每個國家、每個地方都需要為其社會建立人工智能基礎設施(AI infrastructure)。每個政府都應支持為其人民建設人工智能基礎設施，以實現這種能力的民主化。這一點至關重要。人工智能的擴散政策應致力於最大化人工智能的普及，造福每個人。我對此深信不疑。我再舉一個例子：ChatGPT是有史以來最受歡迎的應用程式之一，且僅用了幾年時間就達到這樣的成就。一個應用程式要成為受歡迎的應用程式，必須既實用又易於使用。現在幾乎每個人都會使用ChatGPT。如果你不知道如何使用ChatGPT，你甚至可以直接說：「我不知道怎麼用ChatGPT，要怎麼使用ChatGPT？」

Democratizing AI is very important. It’s absolutely true. It is the reason why every every region, every country, everywhere needs to build AI infrastructure for their society. Every government needs to support building AI infrastructure for their people so that they democratize this capability for them for their people. And so so um That’s That's utterly vital. The AI diffusion policy should be about maximizing diffusion of AI for the benefit of everyone. Democratize AI and I I believe very strongly about that. Um it is also the case. I’ll give you an example: Um one example is the that that ChatGPT is the most popular applications in history, and it only took a couple of years. You can only be a a popular application if it’s useful and easy to use. Everyone is using ChatGPT. If you don’t know how to use ChatGPT, you can even say, “I don’t know how to use ChatGPT. How do I useChatGPT?”

人工智能(AI)會向你解釋它的運作方式，它非常聰明。因此，人工智能不僅民主化了技術，也民主化了知識的獲取。我想分享的第三個例子是我最喜歡的案例之一，這是一個來自Google的實際案例。我記得這是現任Google首席科學家(chief scientist)傑夫·迪恩(Jeff Dean)提到的。他分享了一個關於日本一位黃瓜農夫的故事。這位農夫的兒子，一名高中生，透過下載簡單的人工智能程式，幫助他的母親打造了一個用來分類黃瓜的機器人。分類黃瓜是一項非常困難的任務，因為需要憑藉視覺判斷哪顆黃瓜適合做壽司(sushi cucumber)、哪顆適合賣到雜貨店(grocery cucumber)，以及哪顆適合留來自用烹飪。這些判斷取決於黃瓜的形狀和大小，不同的類型能賣出不同的價格，價格差異可能相當大。那麼，如何教導人工智能完成這樣的任務呢？

It explains to you. It’s so smart. And so AI democratizes technology. It democratizes access to knowledge. The third thing I would say is one of my favorite examples, and this was a Google Google example actually and I remember this was Jeff Dean their their Chief scientist now um said that that was giving an example of a cucumber farmer in Japan and how her son taught an AI to sort cucumbers. Here’s a son,a high school student who who um downloaded some simple AI and helped the mother create a robot to sort cucumbers. Sorting cucumbers is very hard because of her eyes, her sense of which one is a sushi cucumber, which one is a grocery cucumber, and which one is a cucumber that you should keep for yourself to cook for dinner. Depending on which one of the cucumber’s shapes and sizes, they can attract different prices, and the prices could be quite different. So, how do you teach an AI to do that?

這位高中生毫無困難地完成了這項任務。他為他的母親普及了一項技術，坦白說，在此之前，這幾乎是不可能實現的，因為過去需要真人來完成這樣的任務。這是人工智能民主化的另一個完美範例。然而，這並不改變一個事實：我們必須在各地建立人工智能基礎設施(AI infrastructure)。這也是我為什麼如此熱衷於在台灣推動人工智能基礎設施建設的原因。台灣需要擁有自己的AI工廠(AI factory)和人工智能基礎設施，這樣整個產業和社會才能隨著人工智能的進步而發展，並將這項技術交到每個人手中。這真的非常重要。

Well, the high school student had no problem doing that. He democratized a technology for his mother that, quite frankly, until now was impossible to do. You needed a real person to do it. That’s another example of perfect democratization, democratizing. It is the case, though, that it doesn’t change the fact that we have to put AI infrastructure everywhere, which is the reason why I’m so passionate about putting AI infrastructure here in Taiwan. Taiwan needs to have an AI factory and AI infrastructure so that the whole industry and society can advance with AI, putting it in the hands of everyone. So it’s really important.

你為台灣所做的事情充滿熱情和動力，這真的非常重要。現在我們進入第二部分，這是來自一位名叫蔡欣瑜(Cai Xin Yu)的問題。她是國立長沙大學全球傳播與創新科技碩士課程的二年級學生，專攻人文学科(humanities)。她的問題是：許多年輕人，特別是從事人文学科研究的年輕人，正面臨如何將自身專業與人工智能(AI)的未來產業趨勢結合的挑戰。你認為學生們該如何將人工智能技術應用於人文学科和生命科學(life sciences)，以應對未來的產業變革和轉型？

There’s a lot of passion and enthusiasm in the things you have done for Taiwan. So, we move to the second part. This is a question from Cai Xin Yu. She’s a second-year master’s student in the master’s program in Global Communication and Innovation Technology at National Changsha University. She’s from the humanities side. Her question is: Many young people, particularly those in the humanities, are facing the challenge of aligning with future industry trends in AI. So, how do you think students can apply AI technology to the fields of humanities and life sciences to address future industries and transformations?

我的首要建議是：立即開始接觸人工智能(AI)。我每天都使用人工智能進行研究。當我需要進行一些相當複雜的研究時，我會向人工智能描述我的需求，然後它會自動去閱讀大量研究論文和網站內容，幫我整理出所需資訊。

My first and best recommendation is to engage with AI right now. I use AI to do research every day. I describe some fairly complex research I want to do, and my AI goes off and reads a whole bunch of research papers and studies a whole bunch of websites.

人工智能(AI)會進行各種各樣的閱讀，然後為我生成一份非常優雅且易於理解的報告。但更重要的是，它不僅僅是生成一份PDF文件，它實際上學習了所有這些領域的知識。因此，它現在可能成為量子物理學(quantum physics)、數位生物學(digital biology)，甚至是永續能源(sustainable energy)的專家。這真的非常不可思議。它不僅僅是完成了一份研究報告，而是成為了一個我可以隨時與之對話、互動的專家，因為它已經進行了大量的背景研究。對於人文学科(humanities)和生命科學(life sciences)的學生來說，你們現在擁有了一個世界上最好的工具來提升知識、進行互動、對話、頭腦風暴(brainstorm)和研究。這是一個令人難以置信的工具，應該立即開始使用它。

It does all kinds of reading and comes back and generates this report for me that’s really elegant and really simple to understand. But the important thing is, it didn’t just generate a PDF for me. It actually learned all those subjects. So now, it’s an expert in maybe quantum physics, maybe it’s an expert in digital biology, maybe it’s an expert in sustainable energy. It’s incredible. It comes back not just producing research, but now this is an expert that I can talk to all the time and interact with because it’s done a lot of research. So, for the humanities students and life sciences students, you now have the world’s best tool to advance your knowledge, to interact with, to dialogue, to brainstorm, and to do research with. It’s an incredible tool, engage it right away.

我想談談你的領導風格。顯然，在過去20年，當我們還沒有實現人工智能(AI)的重大突破之前，這意味著你需要堅持不懈地追求願景，並與你所合作的龐大團隊保持一致。所以我很好奇，你的核心價值觀是什麼？這些價值觀如何引導你在領導、創新和實現持續成功方面的表現？

I want to talk about your leadership and, obviously, for the past 20 years, before we had this whole big AI breakthrough, it really meant that you needed to persevere to achieve the vision and work with a huge team. So, I’m curious, what are some of your core values, and how do they guide you in leadership, innovation, and achieving continued success?

我的核心理念與我們的使命是一致的，那就是做出貢獻，打造世界上其他人無法做到、但極具挑戰性的成果。如果我們在這些看似不可能的任務上成功，即使成功的機率非常渺茫，只要我們做到了，就能以某種更好的方式改變世界。這是我們所有行動的核心。我們所做的一切都很困難。例如，我們在機器人學(robotics)和人工智能(AI)上已經投入了將近15年，自動駕駛汽車(self-driving cars)研究了10年，量子與經典計算(quantum and classical computing)投入了5年，數位生物學(digital biology)也研究了10年，這樣的例子還有很多。對我來說，選擇那些「幾乎可能」(nearly possible)的目標非常重要，而「幾乎可能」對我來說大約是10年的時間範圍。如果我能在10年內將某項技術推向主流，那麼這就是值得我們全力追求的目標。我喜歡以10年為成功視野，這給了我許多學習和迭代的機會。為了做到這一點，你需要有預見未來的能力，連繫各個線索，並想像10年後的世界會是什麼樣子。對我們這樣的開創性公司(pioneering company)來說，這10年的目標對於成熟的技術公司可能需要15年甚至20年。因此，我們必須活在未來。作為首席執行官(CEO)和公司領導者，我們需要思考什麼是可能的、什麼是值得追求的。此外，我會不斷重新評估自己的理由，持續檢查我的假設。因為我的假設以及我用來制定策略、結論或方向的原則，都是非常根本的東西。我必須不斷重新審視這些假設，確認它們是否基於現在已經改變的事實，或者我用來預測未來的原則是否已經改變。

At the core is our mission. My mission is one and the same: to make a contribution by building things that are incredibly hard to do, that no one else in the world is doing or can do. And if we succeed in the unlikely chance that we succeed, we make a real contribution, then we change the world in some better way. That is at the core of everything we do. Everything we do is hard. That’s why it took 30 years. Robotics and AI, we’ve been working on for almost 15 years; self-driving cars, 10 years; quantum and classical computing, 5 years. The list goes on: digital biology, 10 years. The list of examples of things we’ve been pursuing for a long time is extensive. I think the way to approach that is to choose what’s nearly possible. It’s important that the CEO chooses endeavors that are nearly possible. For me, nearly possible is about 10 years. If I can achieve something and make it mainstream within 10 years, it’s time for us to really go and pursue it. I like the 10-year horizon for success. It gives me many chances to learn and iterate. In order to do that 10 years out, you need to be able to see around corners, connect dots, and have an imagination about what 10 years from now looks like. For us, this is not 10 years for a mature technology company. This is 10 years for a pioneering company. For most companies, that would be like 15 or 20 years away. So, we really have to live in the future. As the CEO, the leaders of our company, we have to reason about what’s possible and what’s worthwhile for us to pursue. Then, after that, you just have to reevaluate your reasons. I double-check my assumptions all the time continuously. Because my assumptions and the principles by which I arrive at certain strategies, conclusions, or directions are really fundamental. I have to go back and reevaluate my assumptions continuously. Were the assumptions I made based on facts that are now different? Have the principles I was using to determine the future changed?

這些原則是否仍然穩固？它們真的是基於第一原理(first principles)的合理推論，還是僅僅是當時的暫時性判斷？我必須不斷重新評估這一切。如果我的信念仍然成立，那麼就沒有理由改變方向，只需繼續堅持追求。

Are they still sound? Are they really sound first principles, or were they only temporary at the time? So, I’ve got to reevaluate all of that continuously. If it’s true, and I believe what I believe, then there’s no reason to change and just keep pursuing it.

最後，如果有什麼我還沒有問到，而你想對台灣以及更廣大的觀眾補充的內容，請分享。

Finally, if there’s anything I haven’t asked and you want to add extra points, especially for your Taiwanese audience and beyond.

來源:https://youtu.be/uRuTdREloUI?si=2h7hRtfnRtirheAc