Name: 人工智能的臨界點在哪裡？ (Where Is the Tipping Point for AI?)
Uploaded: 2025-05-26T23:16:04.000Z
Duration: 20 min 38 s

情態副詞

We have 20 minutes exactly to try and understand from a tech perspective, a tech investor perspective, what AI means.

我們有 20 分鐘的時間，從技術角度、技術投資者角度來理解人工智能的含義。

And I'm still amazed at how many big chief executives that I speak to that say, I know AI.

與我交談過的許多大公司的首席執行官都說，我瞭解人工智能，這讓我感到非常驚訝。

程度副詞

I have many AI people that help me, but I don't fully understand the impact this has on me, my staff, and my company in the next five years.

我有很多人工智能人員幫助我，但我並不完全瞭解這在未來五年對我、我的員工和公司的影響。

亨利，人工智能對你意味著什麼？

Is it difficult to have a hard conviction on where AI goes in the next five years?

是否很難對未來五年人工智能的發展方向有堅定的信念？

Well, first of all, the way I see AI is being able to process an intense amount of data information to be able to direct it towards a very specific question or solution that you're searching for.

首先，在我看來，人工智能能夠處理大量數據資訊，並將其導向您正在搜索的特定問題或解決方案。

And obviously, so many advances have been made in recent times just because we've gotten to the point where our computing power has reached the point where we can crunch loads and data and be able to give you responses in a timely manner.

顯然，近來我們已經取得了許多進步，這只是因為我們的計算能力已經達到了可以對負載和數據進行壓縮，並及時做出響應的程度。

But all of you, if you've seen ChatGBT, there's certain things that you can obviously get an instantaneous answer.

但是在座的各位，如果你們看過 ChatGBT，有些事情顯然可以立即得到答案。

There's some things that even that model is still having to figure out how to get your answer and compute.

有些事情，即使是這種模型，也仍然需要弄清楚如何得到答案並進行計算。

And obviously, high intensive sort of computational needs like graphics, video, et cetera, takes more.

顯然，圖形、視頻等高密集型計算需求需要更多時間。

So even advances today like Google's VO3 is an amazing advance in terms of how we're moving so fast in all dimensions of AI.

是以，即使是像谷歌的 VO3 這樣的進步，也是我們在人工智能各個領域飛速發展的驚人進步。

And Henry, just so that everybody knows, basically Phoenix Holdings focuses on technology, financial services, sports, and media.

而亨利，為了讓大家都知道，鳳凰控股基本上專注於科技、金融服務、體育和媒體。

So do you invest in AI or do you also use it to make the best investment decisions?

那麼，您是投資人工智能，還是利用人工智能做出最佳投資決策？

So we're in a unique situation because we're in an emerging market in Vietnam.

是以，我們的情況很特殊，因為我們身處越南的新興市場。

And obviously, we're trying to keep up with all the sort of major technological trends that are obviously transforming the world.

很顯然，我們正在努力跟上所有正在改變世界的主要技術趨勢。

Vietnam itself is just a fast emerging economy.

越南本身只是一個快速崛起的經濟體。

But at the same time, it sees an opportunity in AI in terms of being able to leapfrog into the next generation of technologies.

但與此同時，它也看到了人工智能的機遇，因為人工智能能夠跨越式地發展新一代技術。

OK, so Chris, basically you're the chief executive signal fire.

好吧，克里斯，基本上你是首席執行官信號火。

And this is a venture capital firm. $1 billion is what you raise of fresh capital to back AI companies.

這是一家風險投資公司。你們籌集了 10 億美元的新資金來支持人工智能公司。

So we are the first venture firm actually built on an AI platform.

是以，我們是第一家真正建立在人工智能平臺上的風險投資公司。

So I founded the firm about 13 years ago.

是以，我在大約 13 年前創辦了這家公司。

And from day one, we've been actually practitioners of using AI.

從第一天起，我們就是使用人工智能的實際踐行者。

So we actually track more data on startups than anyone in the world and use that to track about 650 million people, process where people are spending on different applications, and then use that to target what may be the best opportunities.

是以，我們追蹤的初創企業數據實際上比世界上任何人都多，我們利用這些數據追蹤約 6.5 億人，處理人們在不同應用上的消費情況，然後利用這些數據鎖定最佳機會。

Is it difficult, again, to have conviction on one thing?

再問一次，對一件事充滿信心是否很困難？

Because there's so much and there's so many players.

坦率地說，它們還在不斷變化。

So we've been more concerned about the infrastructure layer, that it's rapidly commoditizing, and that the pricing has been dropping by orders of magnitude each year.

是以，我們更關注基礎設施層，它正在迅速商品化，而且價格每年都在以數量級下降。

So it's a tough time to be at the infrastructure layer, but a very interesting time to be at the application layer, where so much data is sitting in silos with inside every type of organization on the planet, where AI can bring that into a pool that can then, over time, be used to take more action.

是以，對於基礎設施層來說，這是一個艱難的時刻，但對於應用層來說，這是一個非常有趣的時刻，因為在這個星球上的每一種組織內部，都存在著大量的數據孤島，而人工智能可以將這些數據彙集到一個池子裡，然後隨著時間的推移，可以用來採取更多的行動。

And so we see the land grab around aggregating that data and the concept of data gravity and data network effects as a sort of core investment thesis.

是以，我們將圍繞數據聚合、數據引力和數據網絡效應概念的土地爭奪視為一種核心投資理論。

So Mia, how do you use AI first at Multiply?

那麼米婭，你是如何在乘法公司首先使用人工智能的呢？

So basically for the investments that you look at and the companies that you buy.

所以，基本上是針對你所關注的投資項目和你所購買的公司。

And also, after we invest, we move them to our operation alarm, in which we start integrating them into our platform in the sectors that we operate in.

此外，在我們投資之後，我們會將它們轉移到我們的營運報警器中，開始將它們整合到我們的平臺上，在我們營運的行業中進行營運。

At the investment stage, for the last maybe few years, we've been doing legal, financial, technical due diligence, as we should.

在投資階段，也許在過去的幾年裡，我們一直在做法律、財務和技術方面的盡職調查，這是我們應該做的。

But we've recently included AI and cybersecurity that falls under technical due diligence, because it's important for us when we invest in a company to know that this company is optimizing its output and the way it operates.

但我們最近將人工智能和網絡安全納入了技術盡職調查的範疇，因為當我們投資一家公司時，我們必須知道這家公司正在優化其產出和營運方式。

Once the investment is done and moves to our operating group, we use AI across several touch points.

一旦投資完成並轉移到我們的營運團隊，我們就會在多個接觸點上使用人工智能。

First of all, at the board level, we've recently introduced MAI, which is an AI board observer that enhances decision-making, executive decision-making, at the level of the board, and also brings more transparency and compliance to the shareholders.

首先，在董事會層面，我們最近引入了 MAI，這是一個人工智能董事會觀察員，它能增強董事會層面的決策、執行決策，也能為股東帶來更多的透明度和合規性。

We've also integrated multiple AI tools across our financial, legal, recruitment departments.

我們還在財務、法律和招聘部門整合了多種人工智能工具。

And we've seen a lot of efficiency come out of that.

我們也是以看到了很多效率的提高。

At this level, we've counted efficiency in terms of man hour, and we've seen 140,000 man hours saved at this level.

在這個層面上，我們以工時來計算效率，在這個層面上我們已經節省了 14 萬個工時。

But more importantly, so that's for the shared services at the board level, and more importantly, across each vertical, it has become more and more interesting and easier for us to use AI tools in order to really create, extract value and create good savings.

但更重要的是，對於董事會層面的共享服務而言，更重要的是，在各個垂直領域，使用人工智能工具來真正創造、提取價值和節省開支變得越來越有趣，也越來越容易。

In our media vertical, for example, we have four companies.

以我們的媒體垂直業務為例，我們有四家公司。

Now they talk to each other, now they sell each other's  So, with this new inventory, and also with this AI integration, hyper-targeting the client when we really use the data that we have, anonymize it, and use it to hyper-target clients in media, we've created tangible 17% uplift on the EBITDA of this particular vertical.

是以，有了這些新的庫存，再加上人工智能整合，當我們真正使用我們所擁有的數據，將其匿名化，並將其用於媒體的超目標客戶時，我們已經為這一特定垂直行業的息稅折舊攤銷前利潤創造了 17% 的有形提升。

Also, we have adjunct AI's, we have initiatives that empower and actually ask, demand employees to start learning how to better integrate AI tools in their operations.

此外，我們還有輔助人工智能，我們採取了一些措施，授權並要求員工開始學習如何更好地將人工智能工具整合到他們的業務中。

Could I ask each of you actually what AI Utopia looks like?

我能問問你們每個人，人工智能烏托邦到底是什麼樣的嗎？

Again, like in four or five years, I mean, the advancement is so fast.

同樣，四五年後，我的意思是，進步是如此之快。

We don't really know, and we have valuations of DeepSeek, we don't really know where the best innovators come from, or what sticks.

我們並不真正知道，我們也有 DeepSeek 的估值，但我們並不真正知道最好的創新者來自哪裡，也不知道什麼能堅持下來。

We know OpenAI is now looking at possibly doing hardware.

我們知道，OpenAI 現在正在研究是否可能做硬件。

I mean, do you need to have a blueprint of where you think we're in five years to be a good investor?

我的意思是，要想成為一名優秀的投資者，你是否需要對我們五年後的發展有一個藍圖？

Look, just to give some perspective, five years, in terms of AI, we were just talking backstage, ChatGPT just came out about two and a half years ago.

聽著，從人工智能的角度來看，五年時間，我們剛剛在後臺討論過，ChatGPT 兩年半前剛剛問世。

But to all of us, at least I think to many of us who use it, it feels like it's been out forever.

但對我們所有人來說，至少我認為對我們中許多使用它的人來說，感覺就像它已經推出了很久。

It's become an incredibly useful tool on a personal level, but as Samia said, and as Chris has referred to, what's happening already is a lot of functional work that, frankly, is quite consistent and mechanical, is being, AI is being able to sort of bring efficiency to that.

在個人層面上，它已成為一個非常有用的工具，但正如薩米亞所說，也正如克里斯所提到的，已經發生的事情是，很多功能性工作，坦率地說，是相當一致和機械的，正在被人工智能提高效率。

Now, the Utopia is obviously this next level of how does all of this recursive sort of development and learning that AI is able to do, it starts to create what I'll call, well, we're calling it generative AI, but really becoming creative, right?

現在，"烏托邦 "顯然是下一個層次，即人工智能如何進行遞歸式的發展和學習，並開始創造我稱之為 "生成式人工智能 "的東西，但它真的變得很有創造力，對嗎？

I think that's the Utopia that we're all working towards, and as we get there, I think the sort of horizons open up even larger.

我認為這是我們都在努力實現的烏托邦，當我們到達那裡時，我認為視野會變得更加開闊。

I wanna actually ask Chris about valuations and whether some of the AI is overhyped, but before we get to that, I'm gonna put the audience to work.

實際上，我想問克里斯關於估值以及一些人工智能是否被過度誇大的問題，但在問這個問題之前，我要先讓觀眾們開始工作。

There's a poll for you, you can also ask questions if you want to our great panelists, but the poll is basically, have you used an AI model in the past six months at work?

我們為您準備了一個投票，您也可以向我們的專家提問，但投票的基本內容是：在過去六個月中，您在工作中使用過人工智能模型嗎？

So it can either be personally, whether you used it, or whether within your team or the company have used it.

是以，既可以是你個人使用過，也可以是你的團隊或公司內部使用過。

所以第一，不，我是百分之百的人類。

兩個，是的，而且我是公開的。

三個，是的，但我也偷偷用過。

So we'll get that to settle down in a second.

所以，我們馬上就能解決這個問題。

And Chris, I mean, I guess this goes to the question of what do you say to the naysayers who say the technology, A, is overhyped, but also there's just unrealistic valuations out there for startups?

克里斯，我的意思是，我想這就涉及到一個問題，對於那些說技術被過度誇大，而且初創企業估值不切實際的反對者，你怎麼說？

I would say on an absolute basis, as an industry, the potential is unbounded.

我想說的是，作為一個行業，其絕對潛力是無限的。

這是一個令人難以置信的大好機會。

At the micro level of individual companies, there are many, if not the vast majority, are way overhyped.

從單個公司的微觀層面來看，即使不是絕大多數，也有許多公司被過分誇大。

So I think we underestimate the potential impact of AI in the medium to long term, because it's hard to wrap our brains around how pervasive and fast it will impact us.

是以，我認為我們低估了人工智能在中長期內的潛在影響，因為我們很難想象它對我們的影響有多麼普遍和迅速。

But the hope of some of these companies achieving that, it'll be very difficult, and it will take longer in a lot of these areas than people hope.

但是，希望其中一些公司實現這一目標將非常困難，而且在許多領域所需的時間將比人們希望的要長。

But does it change everything, or are there specific sectors that will be disrupted 10 times more?

但是，它是否會改變一切，或者說，是否有一些特定領域會受到十倍於此的破壞？

I think the areas that I'm most excited about are in the office of the CFO, back office, administrative compliance, distribution, all sorts of layers where there's just a lot of painful bureaucracy and process and paperwork.

我認為，我最感興趣的領域是首席財務官辦公室、後臺辦公室、行政合規、分銷等各個層級，在這些層級中，存在著大量令人痛苦的官僚主義、流程和文書工作。

And reducing that to free people up to do higher order activities would be my utopia in this.

減少這種情況，讓人們能夠從事更高級的活動，這就是我的烏托邦。

Remove the drudgery of your job and allow you to do the reasons that you took the job in the first place.

消除你工作中的瑣碎，讓你做你當初接受這份工作的原因。

But I think it'll take a number of years to get there, and the challenge will actually be less technological, I think, than restructuring companies and workflows and jobs and that transformation that has to happen in organizations to unlock it.

但我認為，要實現這一目標還需要若干年的時間，我認為，所面臨的挑戰實際上不是技術問題，而是公司、工作流程和工作崗位的重組問題，以及企業為實現這一目標而必須進行的轉型問題。

Simit, do you see anything being overhyped?

西米特，你覺得有什麼地方被過分誇大了嗎？

Again, it's interesting how you use it, because it is for the tasks that doesn't make you dream when you start a new job.

同樣，你使用它的方式也很有趣，因為它適用於那些在你開始新工作時不會讓你產生夢想的任務。

Actually, I'm training the team to start asking themselves before we recruit anybody, is that a task that can be done by AI?

事實上，我正在培訓團隊，讓他們在招募人員之前先問問自己：這是一項可以由人工智能完成的任務嗎？

或者，你真的需要一個人來做這件事嗎？

For the time being, this applies to very junior tasks.

目前，這適用於非常初級的任務。

On the hyper-evaluation, and it doesn't bother me, I'm not investing in AI companies, I'm investing in solutions, tools, and applications which will enable my team to produce better results, enable our machines to produce better results, and really create more value to our shareholders, which is what I'm here to do.

關於超級評估，我並不在意，我投資的不是人工智能公司，我投資的是解決方案、工具和應用，它們能讓我的團隊取得更好的成果，讓我們的機器取得更好的成果，真正為我們的股東創造更多價值，這就是我在這裡要做的事情。

If you look at how Multiplier's portfolio is constructed, and if you look at several macroeconomic factors happening within an inflationary macro environment, we have a lot of supply chain disruptions, and that costs money on the way we do business.

如果你看看乘法器的投資組合是如何構建的，如果你看看在通脹的宏觀環境下發生的幾種宏觀經濟因素，我們有很多供應鏈中斷的情況，這對我們的經營方式造成了損失。

And in my mind, the only way to really get over this cost and still deliver profitability and double-digit growth to our shareholders is by embracing AI and augmenting the way we do work.

在我看來，要想真正克服這一成本，併為股東帶來盈利和兩位數的增長，唯一的辦法就是擁抱人工智能，增強我們的工作方式。

And we've been doing that at multiple levels, as I said.

正如我所說，我們一直在多個層面開展這項工作。

How do you, so back to the poll, how do you use, do you use AI models personally?

那麼，回到民意調查，您個人如何使用人工智能模型？

Oh yeah, I started using it before Davos this year.

對了，今年達沃斯之前我就開始用了。

My team, it was a model that our tech team developed in-house and of course, at Davos, you meet a lot of people.

我的團隊，這是我們的技術團隊在內部開發的一個模型，當然，在達沃斯，你會遇到很多人。

And this agent, we named it Cora, and it was on WhatsApp.

我們給這個特工起名叫科拉，它在 WhatsApp 上。

And let's say you introduced me to a Mr. Joe from an ex-company.

比方說，你給我介紹了一位前公司的喬先生。

In a second, I would WhatsApp Cora, ask her, give me the bio, and I'll have a glimpse, short summary of the person I'm meeting in order to know how to direct the conversation.

等一下，我會給科拉發 WhatsAppa，問她，把她的履歷給我，我就能對我要見的人有一個一瞥式的簡短概括，以便知道如何引導談話。

所以，這是我的第一個個人作品。

Today, as a board, as I told you, we have the board observer, but also I have my own virtual board.

今天，作為董事會，正如我告訴過你們的，我們有董事會觀察員，但我也有自己的虛擬董事會。

We've contracted a company called Aleria, a very successful AI company in Abu Dhabi.

我們與阿布扎比一家非常成功的人工智能公司 Aleria 簽訂了合同。

字幕列表影片播放

人工智能的臨界點在哪裡？ (Where Is the Tipping Point for AI?)

perspective

sort

relevant

massive