實際影片長度:14:27.000。原文、繁中、雙語可點擊句子跳轉影片。
0:00.000–0:05.340
The AI boom means that data centers are being built at an unprecedented pace around the world
0:05.340–0:09.873
and Isaac, one country in Asia where the build-out is really obvious is Malaysia.
0:10.140–0:11.407
That's right, Mariko.
0:11.407–0:14.273
Malaysia has emerged as an AI hub in Southeast Asia.
0:14.520–0:19.760
The country produces a lot of components that go into chips that help power AI technology.
0:20.220–0:23.753
The country has also been aggressively building data centers.
0:24.080–0:27.547
Now, the economy in Malaysia is expanding thanks to this boom,
0:27.547–0:29.860
but there's some pushback too.
0:30.000–0:35.133
Yeah, and the big question is whether or not those concerns about data center are justified.
0:35.320–0:41.680
And also, how is the pushback against them affecting Southeast Asia's effort to get a slice of the AI boom?
0:45.540–0:49.437
Every application you use, whether it's with AI or without AI,
0:49.437–0:51.160
has to come from somewhere.
0:51.500–0:55.233
It's not just a little computation that's done in the cloud, right?
0:55.233–0:59.300
The cloud is actually hosted in a physical location called a data center.
0:59.300–1:04.767
One of the things that in these data centers they have is that they will have a lot of network cables,
1:04.767–1:09.059
a lot of what they call server racks that is responsible for the computation,
1:09.059–1:12.580
and then also the cooling, the energy requirements on there.
1:12.680–1:16.280
And usually the facilities are managed by very, very few people.
1:16.280–1:19.276
So the reason why that has been going was because
1:19.276–1:22.672
the compute usage has gone up significantly,
1:22.672–1:25.801
probably powered a lot by ChatGPT, Cloud,
1:25.801–1:28.840
or every other AI application out there.
1:28.980–1:33.228
And that is why we need to, because we need data centers and
1:33.228–1:35.440
we need large sources of land,
1:35.660–1:38.527
you're going to see large sources of construction,
1:38.527–1:40.593
there is changes in the environment,
1:40.593–1:44.560
changes in maybe many normal people's daily expenses.
1:44.800–1:47.200
So that's why it's now become a big subject.
1:47.200–1:48.000
Yeah, indeed.
1:48.000–1:50.867
And yeah, because I was looking at, I think in the UK,
1:50.867–1:52.733
there were some protests against
1:52.733–1:55.467
those data centers being built in your backyard,
1:55.467–1:56.933
also in the United States.
1:56.933–1:58.267
But Isaac, here in Asia,
1:58.267–2:02.600
Malaysia kind of stands out in terms of the number of data centers being built.
2:02.700–2:05.960
How did Malaysia become such a hub in terms of this AI boom?
2:06.040–2:08.832
I think the big contributor is back in 2019 when
2:08.832–2:11.480
Singapore implemented this temporary ban.
2:11.620–2:15.220
So developers had to look elsewhere to put their data centers.
2:15.460–2:18.460
And so they looked into neighboring Malaysia, Johor.
2:18.460–2:23.180
Its proximity to Singapore makes it a perfect place to build data centers,
2:23.180–2:24.614
which you can see now,
2:24.614–2:28.120
85% of Malaysia's data centres are located in Johor.
2:28.420–2:31.640
Furthermore, the government put out incentives
2:31.640–2:35.460
that sort of draw in foreign investors to invest in the country.
2:35.840–2:41.420
So Bernard, Isaac just talked about Singapore's restrictions on data centres back in 2019.
2:41.540–2:43.407
Can you just explain why that was?
2:43.407–2:46.540
It's actually not so much a ban, it's kind of a moratorium
2:46.540–2:49.740
because I think the Singapore government is very prudent
2:49.740–2:52.540
about how they're thinking about the electricity
2:52.540–2:55.040
outages that have been going on.
2:55.140–2:57.940
They're mostly caused because of a spike in demand
2:57.940–2:58.740
in computing.
2:58.760–3:01.840
And I think post 2024,
3:02.420–3:05.500
when there was also the US banning China
3:05.500–3:07.060
for having the NVIDIA chips
3:07.060–3:09.060
and a lot of the Chinese founders
3:09.060–3:11.600
from the Chinese tech companies,
3:11.700–3:13.280
they require another place
3:13.280–3:15.013
where they could actually have
3:15.013–3:16.413
a source of good compute.
3:16.540–3:19.140
So that also drives in a lot of foreign director
3:19.140–3:21.473
investments into Malaysia as a result. So
3:21.473–3:24.760
it's a combination of factors. We also talk about, say,
3:24.900–3:30.012
things like legislation changes. Somewhere between 2019 to 2023,
3:30.012–3:32.020
a lot of the Southeast Asia
3:32.020–3:35.487
countries started to have what we call data sovereignty laws.
3:35.487–3:37.153
So data sovereignty laws say,
3:37.153–3:42.353
for example, Malaysia said we need to have something like the data must be hosted in Malaysia.
3:42.740–3:45.207
Indonesia certainly say the same. So it also
3:45.207–3:47.407
prompts the hyperscalers to say, okay,
3:47.407–3:52.340
we probably need in order to allow our businesses to operate in these countries,
3:52.520–3:54.860
we have to set up the data centers.
3:55.040–3:59.460
And that build up, I think, started as of when that period started.
3:59.660–4:00.593
And I have to say,
4:00.593–4:03.660
they don't look that pretty to have in your back garden.
4:03.660–4:05.660
But Isaac, can you explain?
4:06.300–4:09.700
I mean, it's not just geography of where Malaysia was, right?
4:09.700–4:12.967
That all the investments started pouring into the country
4:12.967–4:16.767
because it has been producing a lot of components, tech components.
4:16.860–4:19.870
So how have all these data centers being built?
4:19.870–4:21.800
How has it helped the economy?
4:21.980–4:25.513
And can you just explain the background of Malaysia's economy?
4:25.513–4:29.647
Yeah, like billions of dollars are being poured into Malaysia's economy.
4:29.647–4:32.513
And according to the International Monetary Fund,
4:32.513–4:33.483
this year alone,
4:33.483–4:37.360
it's expected for Malaysia's GDP to grow by at least 4.7%.
4:37.360–4:42.320
Now, the government obviously has a dream of being at the frontier of this AI revolution.
4:42.320–4:46.320
So it's investing a lot. It's offering eligible companies incentives.
4:46.320–4:50.580
For example, some of them could apply for income tax exemptions.
4:51.000–4:54.600
Others could even apply, could get like investment allowances.
4:54.780–4:59.660
So all these incentives are the government's steps to make that dream a reality.
5:00.380–5:03.505
But of course, there has been quite a lot of protests and
5:03.505–5:05.340
you actually spoke to somebody.
5:05.460–5:09.360
So can you just tell us what she told you, what her concerns are?
5:09.780–5:11.513
Yeah, I spoke to Itha Malayana.
5:11.513–5:13.647
She's a 26-year-old English teacher.
5:13.647–5:18.180
So in July, she saw the news that a data center was being proposed in her community.
5:18.560–5:21.700
And she basically just said, no, not in my backyard.
5:22.120–5:26.308
And that concern is because she saw what's happening in the US,
5:26.308–5:28.080
the huge backlash against
5:28.080–5:31.480
data centers, and she didn't want that to be in her community.
5:31.480–5:34.213
So she goes online and she creates this petition.
5:34.460–5:37.593
And what really stood out for me is that in that petition,
5:37.593–5:39.393
she started with a really famous
5:39.393–5:41.907
quote about how when the last tree falls,
5:41.907–5:45.080
people will realize that they can't eat money.
5:45.080–5:46.547
And I asked her about that.
5:46.547–5:51.600
She said that quote basically encapsulates everything that she's standing against the
5:51.600–5:53.020
data centers.
5:53.020–5:54.687
And this is what she has to say.
5:54.687–6:01.060
One of the justifications of data centers is that local industry workers would get trained
6:01.060–6:04.480
and hopefully get to develop this industry in Malaysia.
6:04.480–6:07.238
But I think I've read very briefly on this, but
6:07.238–6:09.560
it seems like that's not going to be the
6:09.560–6:10.227
case at all.
6:10.320–6:13.120
Like they're just going to set up these servers and
6:13.120–6:15.053
then have a few people managing it.
6:15.053–6:18.320
But it's not really going to be like a skill transfer thing.
6:18.320–6:21.053
I could justify using AI in like medicine or like,
6:21.053–6:22.920
I don't know, astrophysics.
6:22.920–6:29.040
But why do we need it to decide what we want for lunch or solve a relationship problem?
6:29.400–6:32.133
Bernard, what do you think about what Aoife said?
6:32.133–6:35.537
I think there are a few parts of what it comes to when
6:35.537–6:37.660
a data center comes online.
6:38.240–6:40.707
So before the data center is actually built,
6:40.707–6:43.840
that actually brings in what we call construction jobs.
6:43.840–6:46.773
Now, once the data center is complete and completed,
6:47.080–6:50.222
then the management of the data center is taken over by,
6:50.222–6:51.520
say, the hyperscaler.
6:51.800–6:54.467
They typically, it's actually very few people,
6:54.467–6:56.000
as she rightfully pointed.
6:56.000–6:57.800
So what happens here is that then
6:57.800–7:00.734
you have a few workers that is actually able to,
7:00.734–7:05.464
if there's any anomalous things happening within the data center,
7:05.464–7:09.180
they will be the person to resolve the issues on that.
7:09.480–7:12.147
But then when you look out to the bigger economy,
7:12.147–7:16.013
right, then you think about what the data center is actually driving.
7:16.013–7:16.947
The data center,
7:16.947–7:21.013
the real focus is actually to drive more what we call the digital economy.
7:21.013–7:23.813
So the digital economy could be in the form of, say,
7:23.920–7:26.040
maybe knowledge-based work.
7:26.160–7:29.900
It could be new kinds of jobs that may be created out of this AI.
7:30.000–7:32.800
I know we are now in what we call a transition stage.
7:32.900–7:39.300
And so the big fear is the job displacement is larger than the job creation.
7:40.000–7:40.520
Yeah.
7:40.780–7:44.460
The only answer to what she's saying is the creation of new jobs.
7:44.620–7:46.687
Is there going to be more than or less?
7:47.120–7:48.720
That question, I don't know.
7:48.720–7:52.800
I would prefer to be much more humble to say I don't know.
7:53.020–7:58.640
But I think we've seen different iterations of technological change.
8:00.000–8:02.260
And so far, every time it happened,
8:02.420–8:04.420
we seem to end up creating more jobs.
8:04.800–8:07.920
If you're just looking at this current time period,
8:08.020–8:09.687
when's the data center built?
8:09.687–8:11.553
She's effectively right on that.
8:11.553–8:13.620
But how the data center creates value
8:13.620–8:16.153
for the digital economy within that country,
8:16.153–8:17.700
that is not so clear to me.
8:17.800–8:19.933
But would that be within that country?
8:19.933–8:23.067
Because couldn't you argue that those new jobs created,
8:23.067–8:26.180
thanks to the AI boom, might be outside of Malaysia?
8:26.280–8:29.200
So that was the whole reason why every country
8:29.200–8:31.400
now has their own Data Sovereignty Act.
8:31.400–8:34.600
Yes, you can host other data from other parts of the world,
8:34.600–8:37.700
but you still have a sustainable holding.
8:38.120–8:40.320
And every business out there who needs,
8:40.380–8:42.647
whether it's a neocloud or hyperscaler,
8:42.647–8:46.580
they have to invest in Malaysia in order to get that part of the economy.
8:46.720–8:48.920
And they have to drive the local economy
8:48.920–8:52.300
because what they don't really want is having everyone,
8:52.560–8:55.427
the local communities holding pitchforks and say,
8:55.427–8:58.227
hey, you know, we want to burn down the data center.
8:58.227–9:00.427
That is also a cause of concern for them.
9:00.600–9:04.267
And I think that stability also becomes part of the conversation.
9:04.267–9:08.540
How has the Malaysian government been speaking about this ambition,
9:08.540–9:11.397
their ambition to be the AI hub, but also
9:11.397–9:14.860
concerns from those local residents over jobs,
9:14.860–9:16.798
but also about the environment.
9:16.798–9:20.460
I previously mentioned that Johor currently houses about 80
9:20.460–9:22.782
to 85% of the data centers in Malaysia. So
9:22.782–9:26.060
in the state of Johor, the government has banned two tiers
9:26.060–9:29.055
of data centers. First is a tier one and tier two.
9:29.055–9:31.900
Those are data centers which consume about 50
9:31.900–9:34.672
million litres of water a day. And
9:34.672–9:39.820
that amount is basically about, it could satisfy the people's
9:39.820–9:43.500
daily drinking needs above 25 million people. So
9:43.500–9:47.000
the government has stopped those data centres
9:47.000–9:48.671
from being built. It has also
9:48.671–9:52.500
introduced penalties where if the data centre operator said that
9:52.500–9:56.692
this is the amount of electricity it's going to use per year,
9:56.692–9:58.620
if it doesn't need 85% of it,
9:58.620–10:00.620
they're going to find the operator.
10:00.800–10:03.600
There's actually a metric to look at data centers.
10:03.600–10:05.867
It's called power usage effectiveness,
10:05.867–10:07.533
what they call the PUE metric.
10:07.533–10:09.720
So if you think about, say,
10:10.020–10:12.240
currently Singapore is mandating
10:12.240–10:15.160
that all data centers must be below
10:15.160–10:17.400
what is called 1.25 PUE.
10:17.920–10:18.420
Okay.
10:18.580–10:20.220
Malaysia, I think it's 1.4.
10:20.560–10:22.760
So there are two aspects of the problem.
10:22.760–10:24.693
So one of them is the water problem.
10:24.693–10:26.120
The water problem is real
10:26.120–10:31.360
if they do not try to recycle the water for the data center's cooling.
10:31.820–10:36.200
So there are actually already known solutions to that.
10:36.400–10:39.600
So there is something called StatePoint Liquid Cooling.
10:39.600–10:42.333
A couple of companies actually started in the US.
10:42.333–10:43.467
But how do they do it?
10:43.467–10:45.467
So basically, you reflux the water.
10:45.467–10:48.200
There is an entire recycling technology on that.
10:48.420–10:51.780
The more concerning part of this is actually energy.
10:52.900–10:55.833
Because of the usage of AI is exponentially growing,
10:55.833–10:58.500
the energy usage is also growing exponentially
10:58.500–11:00.860
and we are actually filling it.
11:01.200–11:05.640
So a lot of it is due to poor demand forecasting of power.
11:05.840–11:08.507
If you think about, look at the region as a whole,
11:08.507–11:11.440
countries, I mean, Singapore is already doing that.
11:11.440–11:13.173
The next data centre you build,
11:13.173–11:15.040
you have to have a certain energy,
11:15.040–11:16.707
that means the energy you use,
11:16.707–11:19.107
some parts of it must come from renewables.
11:19.107–11:21.507
Then as you go to say Malaysia, for example,
11:21.507–11:23.373
then the question is what kind of,
11:23.373–11:25.040
they definitely have energy,
11:25.040–11:26.507
they have petroleum, oil,
11:26.507–11:29.773
so they can hydrocarbon so you can power some part of it then
11:29.773–11:31.440
the question is do they want to
11:31.440–11:32.973
rely on other mixes because
11:32.973–11:36.820
the there is no way now that uh just by using a carbon they are able
11:36.820–11:38.565
to run that they also you also
11:38.565–11:42.420
bring about the increase of climate change as well and malaysia
11:42.420–11:45.582
has the ambition to be that ai hub in southeast asia and
11:45.582–11:47.620
it really wants to achieve that so i
11:47.620–11:52.753
think the government needs to balance between addressing the people's concerns while also
11:52.753–11:57.090
taking realistic steps to make those dreams a reality of being an AI hub.
11:57.090–11:58.180
So the Malaysian
11:58.180–12:04.660
government, the Prime Minister announced that the young people will be able to access a three-month
12:04.660–12:07.580
free subscription to some of the AI models and
12:07.580–12:10.500
that is basically to allow them to explore the
12:10.500–12:15.306
AI models, play with it and even to perhaps boost their creativity and
12:15.306–12:17.460
improve their digital skills.
12:17.460–12:22.520
So the government is taking an active step into making the country embrace AI.
12:22.740–12:26.980
But I guess we'll see how the government addresses the people's concerns.
12:27.760–12:30.795
Will they keep building data centers for now with,
12:30.795–12:33.180
I guess, stricter regulations, Isaac?
12:33.440–12:35.107
I think so, honestly, because
12:35.107–12:38.040
the people's concerns about data centers are valid.
12:38.040–12:38.973
And for example,
12:38.973–12:44.240
you heard from Ether herself that she said that she doesn't want Malaysia to become like the US,
12:44.240–12:47.440
where the people actively push back against data centers
12:47.440–12:50.561
because she's questioning why is data centers, AI,
12:50.561–12:52.460
relevant or even important?
12:52.460–12:56.420
Because we've heard from the AI company CEOs themselves
12:56.420–12:59.353
that they say that this AI revolution is inevitable.
12:59.620–13:01.220
But she's questioning that.
13:01.320–13:02.943
So I think we'll hear more and
13:02.943–13:05.580
more people having similar concerns as Ether.
13:06.000–13:09.700
One interesting example of a country that's in this AI revolution
13:09.700–13:12.560
that's not very much we would think about is China.
13:12.560–13:15.820
so where China is pretty effective
13:15.820–13:17.520
is the energy infrastructure
13:17.520–13:18.653
so they have built up
13:18.653–13:20.587
most of the energy infrastructure
13:20.587–13:21.587
they have built up
13:21.587–13:22.187
most of the
13:22.187–13:23.460
renewable energy mix
13:23.460–13:24.620
what they don't have
13:24.620–13:25.487
is now the chips
13:25.487–13:27.160
and the AI models are lagging
13:27.160–13:28.820
basically 3 to 6 months
13:28.820–13:30.353
depending on who you talk to
13:30.353–13:31.120
on that
13:31.120–13:32.820
but if you look at say
13:32.820–13:34.400
there was a survey done
13:34.400–13:36.420
80% of Chinese
13:36.420–13:38.320
are hopeful about the AI future
13:38.320–13:40.600
and then if you do the same survey
13:40.600–13:42.080
in the US it's only 31%
13:42.080–13:45.080
So the question is, what's the big difference, right?
13:45.080–13:48.300
So I think a lot of it is in whether we're optimistic
13:48.300–13:52.820
that the AI will bring about sort of a growth of a new economy.
13:53.660–13:59.520
And there are some things that I think people take it to the logical extreme.
13:59.840–14:03.860
If AI was to do everything of this, then what should humans do, right?
14:04.180–14:07.920
I think the way I would probably want to moving ahead,
14:07.920–14:12.260
I think that there is actually human creativity is actually infinite.
14:12.580–14:15.647
The demand for human creativity is actually infinite.
14:15.647–14:17.939
And I think for that, because of that,
14:17.939–14:20.760
even if AI can automate most of the things,
14:21.100–14:23.033
there are still things for us to do.
14:23.033–14:25.300
Yeah, really insightful conversation.
14:25.300–14:27.000
Thank you so much for joining me, Bernad and Isaac.
0:00.000–0:05.340
人工智慧熱潮意味著全球正以前所未有的速度建設數據中心
0:05.340–0:09.873
艾薩克,在亞洲,馬來西亞是數據中心建設最為顯著的國家之一。
0:10.140–0:11.407
沒錯,真理子。
0:11.407–0:14.273
馬來西亞已崛起為東南亞的人工智慧樞紐。
0:14.520–0:19.760
該國生產大量用於製造助力量產人工智慧技術的晶片零組件。
0:20.220–0:23.753
該國也積極建設數據中心
0:24.080–0:27.547
現在,得益於這場熱潮,馬來西亞經濟正在擴張,
0:27.547–0:29.860
但也出現了一些反彈。
0:30.000–0:35.133
是的,而最大的問題在於,人們對數據中心的擔憂是否合理。
0:35.320–0:41.680
此外,這種反彈如何影響東南亞在人工智慧熱潮中分一杯羹的努力?
0:45.540–0:49.437
你使用的每個應用程式,無論是否涉及人工智慧,
0:49.437–0:51.160
都必須來自某個地方。
0:51.500–0:55.233
這不僅僅是在雲端進行的一些計算,對吧?
0:55.233–0:59.300
雲端實際上託管在一個稱為數據中心的物理位置。
0:59.300–1:04.767
這些數據中心擁有的東西之一是它們會有大量的網路電纜,
1:04.767–1:09.059
大量被稱為伺服器機櫃的設備,負責進行計算,
1:09.059–1:12.580
以及相關的冷卻系統和能源需求。
1:12.680–1:16.280
通常,這些設施由非常非常少的人管理。
1:16.280–1:19.276
所以之所以會這樣發展,是因為
1:19.276–1:22.672
運算使用量大幅上升
1:22.672–1:25.801
可能主要由 ChatGPT、雲端
1:25.801–1:28.840
或各種其他人工智慧應用程式推動。
1:28.980–1:33.228
這就是為什麼我們需要,因為我們需要數據中心,
1:33.228–1:35.440
並且需要大量的土地,
1:35.660–1:38.527
你會看到大量的建設活動,
1:38.527–1:40.593
環境也會發生變化,
1:40.593–1:44.560
普通人的日常開銷也可能發生變化。
1:44.800–1:47.200
所以這就是為什麼它現在成為一個大議題。
1:47.200–1:48.000
是的,確實如此。
1:48.000–1:50.867
而且是的,因為我在看,我認為是在英國,
1:50.867–1:52.733
有一些抗議活動
1:52.733–1:55.467
反對在你家後院建設數據中心,
1:55.467–1:56.933
在美國也是如此。
1:56.933–1:58.267
但艾薩克,在亞洲,
1:58.267–2:02.600
就建設數據中心的數量而言,馬來西亞顯得尤為突出。
2:02.700–2:05.960
馬來西亞是如何成為這場人工智慧熱潮中的這樣一個樞紐的?
2:06.040–2:08.832
我認為最大的貢獻因素要追溯到 2019 年,當時
2:08.832–2:11.480
新加坡實施了這項臨時禁令。
2:11.620–2:15.220
因此,開發人員必須尋找其他地方來建設他們的數據中心。
2:15.460–2:18.460
於是他們將目光轉向鄰國馬來西亞的柔佛。
2:18.460–2:23.180
它靠近新加坡的地理位置,使其成為建立資料中心的理想地點,
2:23.180–2:24.614
這你可以看到,
2:24.614–2:28.120
馬來西亞85%的資料中心都位於柔佛州。
2:28.420–2:31.640
此外,政府推出了各種誘因
2:31.640–2:35.460
這種情況會吸引外國投資者投資該國。
2:35.840–2:41.420
所以伯納德、艾薩克剛才談到了新加坡在2019年對資料中心實施的限制。
2:41.540–2:43.407
你能解釋一下當時為什麼會這樣嗎?
2:43.407–2:46.540
這實際上與其說是禁令,不如說是一種暫停期
2:46.540–2:49.740
因為我認為新加坡政府對於電力
2:49.740–2:52.540
中斷問題的思考非常謹慎,
2:52.540–2:55.040
這些中斷問題一直持續發生。
2:55.140–2:57.940
這主要是由於運算需求激增所導致
2:57.940–2:58.740
的。
2:58.760–3:01.840
我認為在2024年之後,
3:02.420–3:05.500
當時美國也禁止中國
3:05.500–3:07.060
使用輝達(NVIDIA)晶片,
3:07.060–3:09.060
許多中國科技公司的
3:09.060–3:11.600
來自中國科技公司的,
3:11.700–3:13.280
需要另一個地方
3:13.280–3:15.013
讓他們能夠真正獲得
3:15.013–3:16.413
優質的運算資源。
3:16.540–3:19.140
因此,這也帶動了大量外國直接投資
3:19.140–3:21.473
對馬來西亞的投資。因此
3:21.473–3:24.760
這是多種因素結合的結果。我們也談到,例如
3:24.900–3:30.012
立法方面的改變。在2019年到2023年之間,
3:30.012–3:32.020
東南亞的許多
3:32.020–3:35.487
國家開始實施我們所謂的「數據主權法」。
3:35.487–3:37.153
數據主權法規定,
3:37.153–3:42.353
例如馬來西亞表示,我們需要類似「數據必須託管在馬來西亞」這樣的規定。
3:42.740–3:45.207
印尼也明確表示同樣的看法。因此,這也
3:45.207–3:47.407
促使超大型雲端服務商表示,好吧,
3:47.407–3:52.340
為了讓我們在這些國家的業務能夠正常運作,
3:52.520–3:54.860
我們必須設立資料中心。
3:55.040–3:59.460
我認為這種累積效應是從那個時期開始的。
3:59.660–4:00.593
我必須說,
4:00.593–4:03.660
它們放在你的後花園裡看起來並不太美觀。
4:03.660–4:05.660
但艾薩克,你能解釋一下嗎?
4:06.300–4:09.700
我的意思是,這不僅僅是馬來西亞地理位置的問題,對吧?
4:09.700–4:12.967
所有投資開始湧入該國
4:12.967–4:16.767
是因為它生產了大量的元件,特別是科技元件。
4:16.860–4:19.870
那麼,這些數據中心的建設情況如何呢?
4:19.870–4:21.800
它對經濟有何幫助?
4:21.980–4:25.513
你能否解釋一下馬來西亞經濟的背景?
4:25.513–4:29.647
是的,數十億美元正湧入馬來西亞經濟。
4:29.647–4:32.513
根據國際貨幣基金組織的數據,
4:32.513–4:33.483
僅在今年,
4:33.483–4:37.360
預計馬來西亞的國內生產總值(GDP)將增長至少 4.7%。
4:37.360–4:42.320
現在,政府顯然希望成為這場 AI 革命的領跑者。
4:42.320–4:46.320
因此,政府投入了大量資金,並為符合資格的企業提供激勵措施。
4:46.320–4:50.580
例如,其中一些企業可以申請所得稅豁免。
4:51.000–4:54.600
其他企業甚至可以申請投資補貼。
4:54.780–4:59.660
因此,這些激勵措施是政府實現該願景的步驟。
5:00.380–5:03.505
但當然,已經出現了相當多的抗議活動,
5:03.505–5:05.340
你實際上與某人進行了對話。
5:05.460–5:09.360
那麼,你能告訴我們她說了什麼,她的擔憂是什麼嗎?
5:09.780–5:11.513
是的,我與 Itha Malayana 進行了對話。
5:11.513–5:13.647
她是一位 26 歲的英語老師。
5:13.647–5:18.180
在七月,她看到新聞報導稱,社區內擬建一座數據中心。
5:18.560–5:21.700
她基本上只是說:「不,不要建在我家後院。」
5:22.120–5:26.308
這種擔憂是因為她看到了美國發生的情況,
5:26.308–5:28.080
對大型數據中心所引發的強烈反彈
5:28.080–5:31.480
數據中心產生的巨大反彈,她不希望這發生在她的社區。
5:31.480–5:34.213
於是她上網創建了這份請願書。
5:34.460–5:37.593
讓我印象特別深刻的是,在那份請願書中,
5:37.593–5:39.393
她引用了一句非常著名的
5:39.393–5:41.907
話,大意是當最後一棵樹倒下時,
5:41.907–5:45.080
人們會意識到他們不能吃錢。
5:45.080–5:46.547
我就此問了她。
5:46.547–5:51.600
她說的那句話基本上概括了她所反對的一切
5:51.600–5:53.020
數據中心的一切。
5:53.020–5:54.687
這就是她想說的。
5:54.687–6:01.060
數據中心建設的理由之一是,當地工業從業人員將接受培訓,
6:01.060–6:04.480
並有望在馬來西亞發展該行業。
6:04.480–6:07.238
但我認為我對此只是簡略閱讀過,但
6:07.238–6:09.560
看來情況完全不是這樣。
6:09.560–6:10.227
他們似乎只是設置這些服務器,
6:10.320–6:13.120
就像他們只是要設置這些伺服器並
6:13.120–6:15.053
但這並不真正涉及技能轉移。
6:15.053–6:18.320
我可以為在醫學或類似領域使用 AI 進行辯護,
6:18.320–6:21.053
我可以為在醫學領域使用 AI 辯護,或者說,
6:21.053–6:22.920
我不知道,像是天文物理學。
6:22.920–6:29.040
但為什麼我們需要它來決定午餐想吃什麼,或是解決人際關係問題呢?
6:29.400–6:32.133
伯納德,你對艾奧菲說的話有什麼看法?
6:32.133–6:35.537
我認為這涉及幾個方面,包括
6:35.537–6:37.660
數據中心上線時,
6:38.240–6:40.707
在數據中心實際建成之前,
6:40.707–6:43.840
這實際上會帶來我們所說的建築業就業機會。
6:43.840–6:46.773
現在,一旦數據中心完工並完成,
6:47.080–6:50.222
接著資料中心的營運管理就會由
6:50.222–6:51.520
例如超規模雲端服務商接管。
6:51.800–6:54.467
他們通常,實際上只需要非常少的人,
6:54.467–6:56.000
正如她正確指出的那樣。
6:56.000–6:57.800
所以這裡發生的情況是,然後
6:57.800–7:00.734
你會有一些工人實際上能夠,
7:00.734–7:05.464
如果數據中心內發生任何異常情況,
7:05.464–7:09.180
他們將是負責解決該問題的人。
7:09.480–7:12.147
但當你看更大的經濟體時,
7:12.147–7:16.013
對吧,然後你會思考數據中心實際上在推動什麼。
7:16.013–7:16.947
數據中心,
7:16.947–7:21.013
真正的重點實際上是要推動更多我們所說的數位經濟。
7:21.013–7:23.813
所以數位經濟可能以例如
7:23.920–7:26.040
可能是知識型工作的形式存在。
7:26.160–7:29.900
它可能是由這項人工智慧創造出的新型工作。
7:30.000–7:32.800
我知道我們現在處於我們所說的過渡階段。
7:32.900–7:39.300
因此,最大的擔憂是就業流失大於就業創造。
7:40.000–7:40.520
是的。
7:40.780–7:44.460
對她所說內容的唯一回應是新工作的創造。
7:44.620–7:46.687
會是更多還是更少呢?
7:47.120–7:48.720
這個問題,我不知道。
7:48.720–7:52.800
我更願意謙虛地說我不知道。
7:53.020–7:58.640
但我認為我們已經看到了不同階段技術變革的迭代。
8:00.000–8:02.260
到目前為止,每次發生時,
8:02.420–8:04.420
我們似乎最終都創造了更多工作。
8:04.800–8:07.920
如果你只看當前的這段時間,
8:08.020–8:09.687
數據中心何時建成?
8:09.687–8:11.553
她在這一點上確實是正確的。
8:11.553–8:13.620
但數據中心如何創造價值
8:13.620–8:16.153
對於該國內部的數位經濟,
8:16.153–8:17.700
這對我來說並不明確。
8:17.800–8:19.933
但這會發生在該國內部嗎?
8:19.933–8:23.067
因為你難道不能辯稱,這些因 AI 熱潮而創造的新工作機會,
8:23.067–8:26.180
可能位於馬來西亞境外嗎?
8:26.280–8:29.200
所以這就是為什麼每個國家
8:29.200–8:31.400
現在都制定自己的數據主權法的原因。
8:31.400–8:34.600
是的,你可以託管來自世界其他地區的數據,
8:34.600–8:37.700
但你仍需保持可持續的持有狀態。
8:38.120–8:40.320
而所有需要數據託管的企業,
8:40.380–8:42.647
無論是新雲(neocloud)還是超大型雲端供應商(hyperscaler),
8:42.647–8:46.580
都必須在馬來西亞投資,以獲取該經濟領域的份額。
8:46.720–8:48.920
他們必須推動當地經濟,
8:48.920–8:52.300
因為他們真正不想看到的是,
8:52.560–8:55.427
當地社區手持草叉高呼,
8:55.427–8:58.227
「嘿,你知道嗎,我們想燒毀數據中心。」
8:58.227–9:00.427
這也是他們擔憂的原因之一。
9:00.600–9:04.267
我認為穩定性也成為討論的一部分。
9:04.267–9:08.540
馬來西亞政府對於這項抱負的說法為何
9:08.540–9:11.397
即成為 AI 中心的抱負,但也包括
9:11.397–9:14.860
當地居民對就業的擔憂,
9:14.860–9:16.798
以及對環境的擔憂。
9:16.798–9:20.460
我之前提到,柔佛州目前擁有馬來西亞約 80
9:20.460–9:22.782
至 85% 的數據中心。因此,
9:22.782–9:26.060
在柔佛州,政府禁止建設兩類
9:26.060–9:29.055
數據中心。第一類是一級和二級數據中心。
9:29.055–9:31.900
這些數據中心每天消耗約 50
9:31.900–9:34.672
百萬公升的水。而
9:34.672–9:39.820
這個水量基本上可以滿足超過 2500 萬人的
9:39.820–9:43.500
每日飲水需求超過2500萬人。所以
9:43.500–9:47.000
政府已停止建設這些數據中心。同時,
9:47.000–9:48.671
政府也引入了罰則,如果數據中心運營商聲稱
9:48.671–9:52.500
引入了處罰措施,如果數據中心運營商聲稱
9:52.500–9:56.692
如果他們實際上不需要使用其中的 85%,
9:56.692–9:58.620
政府將會追查運營商。
9:58.620–10:00.620
實際上有一個指標用於評估數據中心。
10:00.800–10:03.600
它被稱為電力使用效率,
10:03.600–10:05.867
也就是所謂的 PUE 指標。
10:05.867–10:07.533
所以,如果你考慮一下,
10:07.533–10:09.720
目前新加坡強制規定
10:10.020–10:12.240
所有數據中心的 PUE 必須低於
10:12.240–10:15.160
所謂的 1.25 PUE。
10:15.160–10:17.400
所謂的1.25 PUE。
10:17.920–10:18.420
好的。
10:18.580–10:20.220
我認為馬來西亞是1.4。
10:20.560–10:22.760
所以這個問題有兩個面向。
10:22.760–10:24.693
其中一個是水資源問題。
10:24.693–10:26.120
如果他們不嘗試為資料中心冷卻系統回收用水,
10:26.120–10:31.360
如果他們不嘗試為數據中心的冷卻系統回收用水。
10:31.820–10:36.200
因此,其實已有已知的解決方案。
10:36.400–10:39.600
有一種稱為StatePoint的液體冷卻技術。
10:39.600–10:42.333
實際上,有幾家公司在美國開始發展這項技術。
10:42.333–10:43.467
但他們是如何做到的呢?
10:43.467–10:45.467
基本上,就是將水進行回流。
10:45.467–10:48.200
這方面已有完整的回收技術。
10:48.420–10:51.780
更令人擔憂的部分其實是能源。
10:52.900–10:55.833
由於AI的使用量呈指數級增長,
10:55.833–10:58.500
能源消耗也呈指數級增長,
10:58.500–11:00.860
我們實際上正在填補這個缺口。
11:01.200–11:05.640
這其中很大一部分是因為電力需求預測不佳。
11:05.840–11:08.507
如果你從整體區域的角度來看,
11:08.507–11:11.440
我指的是國家,新加坡已經在這樣做了。
11:11.440–11:13.173
你接下來建造的資料中心,
11:13.173–11:15.040
必須具備一定的能源條件,
11:15.040–11:16.707
這意味著你所使用的能源中,
11:16.707–11:19.107
其中一部分必須來自可再生能源。
11:19.107–11:21.507
然後,當你考慮到例如馬來西亞時,
11:21.507–11:23.373
問題就變成了什麼樣的能源組合,
11:23.373–11:25.040
他們肯定擁有能源,
11:25.040–11:26.507
他們擁有石油、原油,
11:26.507–11:29.773
因此他們可以使用碳氫化合物來為部分設施供電,
11:29.773–11:31.440
問題是他們是否願意
11:31.440–11:32.973
依賴其他能源組合,因為
11:32.973–11:36.820
目前沒有任何方法可以僅憑使用碳基能源就
11:36.820–11:38.565
能夠運行,而且這也會
11:38.565–11:42.420
也帶來氣候變遷的加劇,而馬來西亞
11:42.420–11:45.582
則有成為東南亞AI樞紐的抱負,
11:45.582–11:47.620
並且確實希望達成此目標,所以我
11:47.620–11:52.753
我認為政府需要在回應民眾擔憂與採取務實步驟實現成為人工智慧樞紐的夢想之間取得平衡
11:52.753–11:57.090
採取務實的步驟,將成為AI樞紐的夢想化為現實。
11:57.090–11:58.180
因此,馬來西亞
11:58.180–12:04.660
政府,首相宣布年輕人將能獲得三個月的
12:04.660–12:07.580
部分AI模型的訂閱服務,並
12:07.580–12:10.500
這基本上是為了讓他們能夠探索
12:10.500–12:15.306
AI模型,進行體驗,甚至可能激發他們的創造力並
12:15.306–12:17.460
提升他們的數位技能。
12:17.460–12:22.520
因此,政府正採取積極措施,讓國家擁抱AI。
12:22.740–12:26.980
但我猜我們會看到政府如何處理民眾的擔憂。
12:27.760–12:30.795
他們目前是否會繼續建設數據中心,
12:30.795–12:33.180
我認為,艾薩克,這需要更嚴格的監管?
12:33.440–12:35.107
老實說,我認為會如此,因為
12:35.107–12:38.040
民眾對數據中心的擔憂是合理的。
12:38.040–12:38.973
舉例來說
12:38.973–12:44.240
你聽過艾瑟本人說過,她表示不希望馬來西亞變成像美國那樣,
12:44.240–12:47.440
人們積極抵制數據中心的地方
12:47.440–12:50.561
因為她質疑數據中心、AI
12:50.561–12:52.460
為何相關甚至重要?
12:52.460–12:56.420
因為我們聽過AI公司執行長本人說
12:56.420–12:59.353
他們表示這場AI革命是不可避免的。
12:59.620–13:01.220
但她對此提出質疑。
13:01.320–13:02.943
所以我想我們會聽到越來越多
13:02.943–13:05.580
像艾瑟一樣有類似擔憂的人。
13:06.000–13:09.700
一個有趣的例子是,在這次AI革命中
13:09.700–13:12.560
一個我們不太會想到的國家是中國。
13:12.560–13:15.820
中國相當有效的地方
13:15.820–13:17.520
在於能源基礎設施
13:17.520–13:18.653
他們已經建立
13:18.653–13:20.587
大部分的能源基礎設施
13:20.587–13:21.587
他們已經建立
13:21.587–13:22.187
大部分的
13:22.187–13:23.460
可再生能源組合
13:23.460–13:24.620
他們現在缺乏的是
13:24.620–13:25.487
現在是晶片
13:25.487–13:27.160
而AI模型落後
13:27.160–13:28.820
基本上落後3到6個月
13:28.820–13:30.353
這取決於你問的是誰
13:30.353–13:31.120
關於這一點
13:31.120–13:32.820
但如果你看看說
13:32.820–13:34.400
有一項調查顯示
13:34.400–13:36.420
80%的中國人
13:36.420–13:38.320
對AI未來充滿希望
13:38.320–13:40.600
然後如果你進行同樣的調查
13:40.600–13:42.080
在美國只有31%
13:42.080–13:45.080
所以問題是,到底有什麼巨大的差異,對吧?
13:45.080–13:48.300
我認為這很大程度上取決於我們是否樂觀
13:48.300–13:52.820
認為 AI 將帶來一種新經濟體系的增長。
13:53.660–13:59.520
我認為有些人會將這一點推演到邏輯極端。
13:59.840–14:03.860
如果 AI 能完成所有這些工作,那麼人類該做什麼呢,對吧?
14:04.180–14:07.920
我認為我未來可能希望採取的方式是,
14:07.920–14:12.260
我認為人類的創造力其實是無限的。
14:12.580–14:15.647
人類創造力的需求其實是無限的。
14:15.647–14:17.939
我認為正因如此
14:17.939–14:20.760
即使 AI 能自動化大多數事情,
14:21.100–14:23.033
仍然有我們要做的事情。
14:23.033–14:25.300
是的,非常有洞察力的對話。
14:25.300–14:27.000
非常感謝 Bernad 和 Isaac 來與我交流。
0:00.000–0:05.340
The AI boom means that data centers are being built at an unprecedented pace around the world
人工智慧熱潮意味著全球正以前所未有的速度建設數據中心
0:05.340–0:09.873
and Isaac, one country in Asia where the build-out is really obvious is Malaysia.
艾薩克,在亞洲,馬來西亞是數據中心建設最為顯著的國家之一。
0:10.140–0:11.407
That's right, Mariko.
沒錯,真理子。
0:11.407–0:14.273
Malaysia has emerged as an AI hub in Southeast Asia.
馬來西亞已崛起為東南亞的人工智慧樞紐。
0:14.520–0:19.760
The country produces a lot of components that go into chips that help power AI technology.
該國生產大量用於製造助力量產人工智慧技術的晶片零組件。
0:20.220–0:23.753
The country has also been aggressively building data centers.
該國也積極建設數據中心
0:24.080–0:27.547
Now, the economy in Malaysia is expanding thanks to this boom,
現在,得益於這場熱潮,馬來西亞經濟正在擴張,
0:27.547–0:29.860
but there's some pushback too.
但也出現了一些反彈。
0:30.000–0:35.133
Yeah, and the big question is whether or not those concerns about data center are justified.
是的,而最大的問題在於,人們對數據中心的擔憂是否合理。
0:35.320–0:41.680
And also, how is the pushback against them affecting Southeast Asia's effort to get a slice of the AI boom?
此外,這種反彈如何影響東南亞在人工智慧熱潮中分一杯羹的努力?
0:45.540–0:49.437
Every application you use, whether it's with AI or without AI,
你使用的每個應用程式,無論是否涉及人工智慧,
0:49.437–0:51.160
has to come from somewhere.
都必須來自某個地方。
0:51.500–0:55.233
It's not just a little computation that's done in the cloud, right?
這不僅僅是在雲端進行的一些計算,對吧?
0:55.233–0:59.300
The cloud is actually hosted in a physical location called a data center.
雲端實際上託管在一個稱為數據中心的物理位置。
0:59.300–1:04.767
One of the things that in these data centers they have is that they will have a lot of network cables,
這些數據中心擁有的東西之一是它們會有大量的網路電纜,
1:04.767–1:09.059
a lot of what they call server racks that is responsible for the computation,
大量被稱為伺服器機櫃的設備,負責進行計算,
1:09.059–1:12.580
and then also the cooling, the energy requirements on there.
以及相關的冷卻系統和能源需求。
1:12.680–1:16.280
And usually the facilities are managed by very, very few people.
通常,這些設施由非常非常少的人管理。
1:16.280–1:19.276
So the reason why that has been going was because
所以之所以會這樣發展,是因為
1:19.276–1:22.672
the compute usage has gone up significantly,
運算使用量大幅上升
1:22.672–1:25.801
probably powered a lot by ChatGPT, Cloud,
可能主要由 ChatGPT、雲端
1:25.801–1:28.840
or every other AI application out there.
或各種其他人工智慧應用程式推動。
1:28.980–1:33.228
And that is why we need to, because we need data centers and
這就是為什麼我們需要,因為我們需要數據中心,
1:33.228–1:35.440
we need large sources of land,
並且需要大量的土地,
1:35.660–1:38.527
you're going to see large sources of construction,
你會看到大量的建設活動,
1:38.527–1:40.593
there is changes in the environment,
環境也會發生變化,
1:40.593–1:44.560
changes in maybe many normal people's daily expenses.
普通人的日常開銷也可能發生變化。
1:44.800–1:47.200
So that's why it's now become a big subject.
所以這就是為什麼它現在成為一個大議題。
1:47.200–1:48.000
Yeah, indeed.
是的,確實如此。
1:48.000–1:50.867
And yeah, because I was looking at, I think in the UK,
而且是的,因為我在看,我認為是在英國,
1:50.867–1:52.733
there were some protests against
有一些抗議活動
1:52.733–1:55.467
those data centers being built in your backyard,
反對在你家後院建設數據中心,
1:55.467–1:56.933
also in the United States.
在美國也是如此。
1:56.933–1:58.267
But Isaac, here in Asia,
但艾薩克,在亞洲,
1:58.267–2:02.600
Malaysia kind of stands out in terms of the number of data centers being built.
就建設數據中心的數量而言,馬來西亞顯得尤為突出。
2:02.700–2:05.960
How did Malaysia become such a hub in terms of this AI boom?
馬來西亞是如何成為這場人工智慧熱潮中的這樣一個樞紐的?
2:06.040–2:08.832
I think the big contributor is back in 2019 when
我認為最大的貢獻因素要追溯到 2019 年,當時
2:08.832–2:11.480
Singapore implemented this temporary ban.
新加坡實施了這項臨時禁令。
2:11.620–2:15.220
So developers had to look elsewhere to put their data centers.
因此,開發人員必須尋找其他地方來建設他們的數據中心。
2:15.460–2:18.460
And so they looked into neighboring Malaysia, Johor.
於是他們將目光轉向鄰國馬來西亞的柔佛。
2:18.460–2:23.180
Its proximity to Singapore makes it a perfect place to build data centers,
它靠近新加坡的地理位置,使其成為建立資料中心的理想地點,
2:23.180–2:24.614
which you can see now,
這你可以看到,
2:24.614–2:28.120
85% of Malaysia's data centres are located in Johor.
馬來西亞85%的資料中心都位於柔佛州。
2:28.420–2:31.640
Furthermore, the government put out incentives
此外,政府推出了各種誘因
2:31.640–2:35.460
that sort of draw in foreign investors to invest in the country.
這種情況會吸引外國投資者投資該國。
2:35.840–2:41.420
So Bernard, Isaac just talked about Singapore's restrictions on data centres back in 2019.
所以伯納德、艾薩克剛才談到了新加坡在2019年對資料中心實施的限制。
2:41.540–2:43.407
Can you just explain why that was?
你能解釋一下當時為什麼會這樣嗎?
2:43.407–2:46.540
It's actually not so much a ban, it's kind of a moratorium
這實際上與其說是禁令,不如說是一種暫停期
2:46.540–2:49.740
because I think the Singapore government is very prudent
因為我認為新加坡政府對於電力
2:49.740–2:52.540
about how they're thinking about the electricity
中斷問題的思考非常謹慎,
2:52.540–2:55.040
outages that have been going on.
這些中斷問題一直持續發生。
2:55.140–2:57.940
They're mostly caused because of a spike in demand
這主要是由於運算需求激增所導致
2:57.940–2:58.740
in computing.
的。
2:58.760–3:01.840
And I think post 2024,
我認為在2024年之後,
3:02.420–3:05.500
when there was also the US banning China
當時美國也禁止中國
3:05.500–3:07.060
for having the NVIDIA chips
使用輝達(NVIDIA)晶片,
3:07.060–3:09.060
and a lot of the Chinese founders
許多中國科技公司的
3:09.060–3:11.600
from the Chinese tech companies,
來自中國科技公司的,
3:11.700–3:13.280
they require another place
需要另一個地方
3:13.280–3:15.013
where they could actually have
讓他們能夠真正獲得
3:15.013–3:16.413
a source of good compute.
優質的運算資源。
3:16.540–3:19.140
So that also drives in a lot of foreign director
因此,這也帶動了大量外國直接投資
3:19.140–3:21.473
investments into Malaysia as a result. So
對馬來西亞的投資。因此
3:21.473–3:24.760
it's a combination of factors. We also talk about, say,
這是多種因素結合的結果。我們也談到,例如
3:24.900–3:30.012
things like legislation changes. Somewhere between 2019 to 2023,
立法方面的改變。在2019年到2023年之間,
3:30.012–3:32.020
a lot of the Southeast Asia
東南亞的許多
3:32.020–3:35.487
countries started to have what we call data sovereignty laws.
國家開始實施我們所謂的「數據主權法」。
3:35.487–3:37.153
So data sovereignty laws say,
數據主權法規定,
3:37.153–3:42.353
for example, Malaysia said we need to have something like the data must be hosted in Malaysia.
例如馬來西亞表示,我們需要類似「數據必須託管在馬來西亞」這樣的規定。
3:42.740–3:45.207
Indonesia certainly say the same. So it also
印尼也明確表示同樣的看法。因此,這也
3:45.207–3:47.407
prompts the hyperscalers to say, okay,
促使超大型雲端服務商表示,好吧,
3:47.407–3:52.340
we probably need in order to allow our businesses to operate in these countries,
為了讓我們在這些國家的業務能夠正常運作,
3:52.520–3:54.860
we have to set up the data centers.
我們必須設立資料中心。
3:55.040–3:59.460
And that build up, I think, started as of when that period started.
我認為這種累積效應是從那個時期開始的。
3:59.660–4:00.593
And I have to say,
我必須說,
4:00.593–4:03.660
they don't look that pretty to have in your back garden.
它們放在你的後花園裡看起來並不太美觀。
4:03.660–4:05.660
But Isaac, can you explain?
但艾薩克,你能解釋一下嗎?
4:06.300–4:09.700
I mean, it's not just geography of where Malaysia was, right?
我的意思是,這不僅僅是馬來西亞地理位置的問題,對吧?
4:09.700–4:12.967
That all the investments started pouring into the country
所有投資開始湧入該國
4:12.967–4:16.767
because it has been producing a lot of components, tech components.
是因為它生產了大量的元件,特別是科技元件。
4:16.860–4:19.870
So how have all these data centers being built?
那麼,這些數據中心的建設情況如何呢?
4:19.870–4:21.800
How has it helped the economy?
它對經濟有何幫助?
4:21.980–4:25.513
And can you just explain the background of Malaysia's economy?
你能否解釋一下馬來西亞經濟的背景?
4:25.513–4:29.647
Yeah, like billions of dollars are being poured into Malaysia's economy.
是的,數十億美元正湧入馬來西亞經濟。
4:29.647–4:32.513
And according to the International Monetary Fund,
根據國際貨幣基金組織的數據,
4:32.513–4:33.483
this year alone,
僅在今年,
4:33.483–4:37.360
it's expected for Malaysia's GDP to grow by at least 4.7%.
預計馬來西亞的國內生產總值(GDP)將增長至少 4.7%。
4:37.360–4:42.320
Now, the government obviously has a dream of being at the frontier of this AI revolution.
現在,政府顯然希望成為這場 AI 革命的領跑者。
4:42.320–4:46.320
So it's investing a lot. It's offering eligible companies incentives.
因此,政府投入了大量資金,並為符合資格的企業提供激勵措施。
4:46.320–4:50.580
For example, some of them could apply for income tax exemptions.
例如,其中一些企業可以申請所得稅豁免。
4:51.000–4:54.600
Others could even apply, could get like investment allowances.
其他企業甚至可以申請投資補貼。
4:54.780–4:59.660
So all these incentives are the government's steps to make that dream a reality.
因此,這些激勵措施是政府實現該願景的步驟。
5:00.380–5:03.505
But of course, there has been quite a lot of protests and
但當然,已經出現了相當多的抗議活動,
5:03.505–5:05.340
you actually spoke to somebody.
你實際上與某人進行了對話。
5:05.460–5:09.360
So can you just tell us what she told you, what her concerns are?
那麼,你能告訴我們她說了什麼,她的擔憂是什麼嗎?
5:09.780–5:11.513
Yeah, I spoke to Itha Malayana.
是的,我與 Itha Malayana 進行了對話。
5:11.513–5:13.647
She's a 26-year-old English teacher.
她是一位 26 歲的英語老師。
5:13.647–5:18.180
So in July, she saw the news that a data center was being proposed in her community.
在七月,她看到新聞報導稱,社區內擬建一座數據中心。
5:18.560–5:21.700
And she basically just said, no, not in my backyard.
她基本上只是說:「不,不要建在我家後院。」
5:22.120–5:26.308
And that concern is because she saw what's happening in the US,
這種擔憂是因為她看到了美國發生的情況,
5:26.308–5:28.080
the huge backlash against
對大型數據中心所引發的強烈反彈
5:28.080–5:31.480
data centers, and she didn't want that to be in her community.
數據中心產生的巨大反彈,她不希望這發生在她的社區。
5:31.480–5:34.213
So she goes online and she creates this petition.
於是她上網創建了這份請願書。
5:34.460–5:37.593
And what really stood out for me is that in that petition,
讓我印象特別深刻的是,在那份請願書中,
5:37.593–5:39.393
she started with a really famous
她引用了一句非常著名的
5:39.393–5:41.907
quote about how when the last tree falls,
話,大意是當最後一棵樹倒下時,
5:41.907–5:45.080
people will realize that they can't eat money.
人們會意識到他們不能吃錢。
5:45.080–5:46.547
And I asked her about that.
我就此問了她。
5:46.547–5:51.600
She said that quote basically encapsulates everything that she's standing against the
她說的那句話基本上概括了她所反對的一切
5:51.600–5:53.020
data centers.
數據中心的一切。
5:53.020–5:54.687
And this is what she has to say.
這就是她想說的。
5:54.687–6:01.060
One of the justifications of data centers is that local industry workers would get trained
數據中心建設的理由之一是,當地工業從業人員將接受培訓,
6:01.060–6:04.480
and hopefully get to develop this industry in Malaysia.
並有望在馬來西亞發展該行業。
6:04.480–6:07.238
But I think I've read very briefly on this, but
但我認為我對此只是簡略閱讀過,但
6:07.238–6:09.560
it seems like that's not going to be the
看來情況完全不是這樣。
6:09.560–6:10.227
case at all.
他們似乎只是設置這些服務器,
6:10.320–6:13.120
Like they're just going to set up these servers and
就像他們只是要設置這些伺服器並
6:13.120–6:15.053
then have a few people managing it.
但這並不真正涉及技能轉移。
6:15.053–6:18.320
But it's not really going to be like a skill transfer thing.
我可以為在醫學或類似領域使用 AI 進行辯護,
6:18.320–6:21.053
I could justify using AI in like medicine or like,
我可以為在醫學領域使用 AI 辯護,或者說,
6:21.053–6:22.920
I don't know, astrophysics.
我不知道,像是天文物理學。
6:22.920–6:29.040
But why do we need it to decide what we want for lunch or solve a relationship problem?
但為什麼我們需要它來決定午餐想吃什麼,或是解決人際關係問題呢?
6:29.400–6:32.133
Bernard, what do you think about what Aoife said?
伯納德,你對艾奧菲說的話有什麼看法?
6:32.133–6:35.537
I think there are a few parts of what it comes to when
我認為這涉及幾個方面,包括
6:35.537–6:37.660
a data center comes online.
數據中心上線時,
6:38.240–6:40.707
So before the data center is actually built,
在數據中心實際建成之前,
6:40.707–6:43.840
that actually brings in what we call construction jobs.
這實際上會帶來我們所說的建築業就業機會。
6:43.840–6:46.773
Now, once the data center is complete and completed,
現在,一旦數據中心完工並完成,
6:47.080–6:50.222
then the management of the data center is taken over by,
接著資料中心的營運管理就會由
6:50.222–6:51.520
say, the hyperscaler.
例如超規模雲端服務商接管。
6:51.800–6:54.467
They typically, it's actually very few people,
他們通常,實際上只需要非常少的人,
6:54.467–6:56.000
as she rightfully pointed.
正如她正確指出的那樣。
6:56.000–6:57.800
So what happens here is that then
所以這裡發生的情況是,然後
6:57.800–7:00.734
you have a few workers that is actually able to,
你會有一些工人實際上能夠,
7:00.734–7:05.464
if there's any anomalous things happening within the data center,
如果數據中心內發生任何異常情況,
7:05.464–7:09.180
they will be the person to resolve the issues on that.
他們將是負責解決該問題的人。
7:09.480–7:12.147
But then when you look out to the bigger economy,
但當你看更大的經濟體時,
7:12.147–7:16.013
right, then you think about what the data center is actually driving.
對吧,然後你會思考數據中心實際上在推動什麼。
7:16.013–7:16.947
The data center,
數據中心,
7:16.947–7:21.013
the real focus is actually to drive more what we call the digital economy.
真正的重點實際上是要推動更多我們所說的數位經濟。
7:21.013–7:23.813
So the digital economy could be in the form of, say,
所以數位經濟可能以例如
7:23.920–7:26.040
maybe knowledge-based work.
可能是知識型工作的形式存在。
7:26.160–7:29.900
It could be new kinds of jobs that may be created out of this AI.
它可能是由這項人工智慧創造出的新型工作。
7:30.000–7:32.800
I know we are now in what we call a transition stage.
我知道我們現在處於我們所說的過渡階段。
7:32.900–7:39.300
And so the big fear is the job displacement is larger than the job creation.
因此,最大的擔憂是就業流失大於就業創造。
7:40.000–7:40.520
Yeah.
是的。
7:40.780–7:44.460
The only answer to what she's saying is the creation of new jobs.
對她所說內容的唯一回應是新工作的創造。
7:44.620–7:46.687
Is there going to be more than or less?
會是更多還是更少呢?
7:47.120–7:48.720
That question, I don't know.
這個問題,我不知道。
7:48.720–7:52.800
I would prefer to be much more humble to say I don't know.
我更願意謙虛地說我不知道。
7:53.020–7:58.640
But I think we've seen different iterations of technological change.
但我認為我們已經看到了不同階段技術變革的迭代。
8:00.000–8:02.260
And so far, every time it happened,
到目前為止,每次發生時,
8:02.420–8:04.420
we seem to end up creating more jobs.
我們似乎最終都創造了更多工作。
8:04.800–8:07.920
If you're just looking at this current time period,
如果你只看當前的這段時間,
8:08.020–8:09.687
when's the data center built?
數據中心何時建成?
8:09.687–8:11.553
She's effectively right on that.
她在這一點上確實是正確的。
8:11.553–8:13.620
But how the data center creates value
但數據中心如何創造價值
8:13.620–8:16.153
for the digital economy within that country,
對於該國內部的數位經濟,
8:16.153–8:17.700
that is not so clear to me.
這對我來說並不明確。
8:17.800–8:19.933
But would that be within that country?
但這會發生在該國內部嗎?
8:19.933–8:23.067
Because couldn't you argue that those new jobs created,
因為你難道不能辯稱,這些因 AI 熱潮而創造的新工作機會,
8:23.067–8:26.180
thanks to the AI boom, might be outside of Malaysia?
可能位於馬來西亞境外嗎?
8:26.280–8:29.200
So that was the whole reason why every country
所以這就是為什麼每個國家
8:29.200–8:31.400
now has their own Data Sovereignty Act.
現在都制定自己的數據主權法的原因。
8:31.400–8:34.600
Yes, you can host other data from other parts of the world,
是的,你可以託管來自世界其他地區的數據,
8:34.600–8:37.700
but you still have a sustainable holding.
但你仍需保持可持續的持有狀態。
8:38.120–8:40.320
And every business out there who needs,
而所有需要數據託管的企業,
8:40.380–8:42.647
whether it's a neocloud or hyperscaler,
無論是新雲(neocloud)還是超大型雲端供應商(hyperscaler),
8:42.647–8:46.580
they have to invest in Malaysia in order to get that part of the economy.
都必須在馬來西亞投資,以獲取該經濟領域的份額。
8:46.720–8:48.920
And they have to drive the local economy
他們必須推動當地經濟,
8:48.920–8:52.300
because what they don't really want is having everyone,
因為他們真正不想看到的是,
8:52.560–8:55.427
the local communities holding pitchforks and say,
當地社區手持草叉高呼,
8:55.427–8:58.227
hey, you know, we want to burn down the data center.
「嘿,你知道嗎,我們想燒毀數據中心。」
8:58.227–9:00.427
That is also a cause of concern for them.
這也是他們擔憂的原因之一。
9:00.600–9:04.267
And I think that stability also becomes part of the conversation.
我認為穩定性也成為討論的一部分。
9:04.267–9:08.540
How has the Malaysian government been speaking about this ambition,
馬來西亞政府對於這項抱負的說法為何
9:08.540–9:11.397
their ambition to be the AI hub, but also
即成為 AI 中心的抱負,但也包括
9:11.397–9:14.860
concerns from those local residents over jobs,
當地居民對就業的擔憂,
9:14.860–9:16.798
but also about the environment.
以及對環境的擔憂。
9:16.798–9:20.460
I previously mentioned that Johor currently houses about 80
我之前提到,柔佛州目前擁有馬來西亞約 80
9:20.460–9:22.782
to 85% of the data centers in Malaysia. So
至 85% 的數據中心。因此,
9:22.782–9:26.060
in the state of Johor, the government has banned two tiers
在柔佛州,政府禁止建設兩類
9:26.060–9:29.055
of data centers. First is a tier one and tier two.
數據中心。第一類是一級和二級數據中心。
9:29.055–9:31.900
Those are data centers which consume about 50
這些數據中心每天消耗約 50
9:31.900–9:34.672
million litres of water a day. And
百萬公升的水。而
9:34.672–9:39.820
that amount is basically about, it could satisfy the people's
這個水量基本上可以滿足超過 2500 萬人的
9:39.820–9:43.500
daily drinking needs above 25 million people. So
每日飲水需求超過2500萬人。所以
9:43.500–9:47.000
the government has stopped those data centres
政府已停止建設這些數據中心。同時,
9:47.000–9:48.671
from being built. It has also
政府也引入了罰則,如果數據中心運營商聲稱
9:48.671–9:52.500
introduced penalties where if the data centre operator said that
引入了處罰措施,如果數據中心運營商聲稱
9:52.500–9:56.692
this is the amount of electricity it's going to use per year,
如果他們實際上不需要使用其中的 85%,
9:56.692–9:58.620
if it doesn't need 85% of it,
政府將會追查運營商。
9:58.620–10:00.620
they're going to find the operator.
實際上有一個指標用於評估數據中心。
10:00.800–10:03.600
There's actually a metric to look at data centers.
它被稱為電力使用效率,
10:03.600–10:05.867
It's called power usage effectiveness,
也就是所謂的 PUE 指標。
10:05.867–10:07.533
what they call the PUE metric.
所以,如果你考慮一下,
10:07.533–10:09.720
So if you think about, say,
目前新加坡強制規定
10:10.020–10:12.240
currently Singapore is mandating
所有數據中心的 PUE 必須低於
10:12.240–10:15.160
that all data centers must be below
所謂的 1.25 PUE。
10:15.160–10:17.400
what is called 1.25 PUE.
所謂的1.25 PUE。
10:17.920–10:18.420
Okay.
好的。
10:18.580–10:20.220
Malaysia, I think it's 1.4.
我認為馬來西亞是1.4。
10:20.560–10:22.760
So there are two aspects of the problem.
所以這個問題有兩個面向。
10:22.760–10:24.693
So one of them is the water problem.
其中一個是水資源問題。
10:24.693–10:26.120
The water problem is real
如果他們不嘗試為資料中心冷卻系統回收用水,
10:26.120–10:31.360
if they do not try to recycle the water for the data center's cooling.
如果他們不嘗試為數據中心的冷卻系統回收用水。
10:31.820–10:36.200
So there are actually already known solutions to that.
因此,其實已有已知的解決方案。
10:36.400–10:39.600
So there is something called StatePoint Liquid Cooling.
有一種稱為StatePoint的液體冷卻技術。
10:39.600–10:42.333
A couple of companies actually started in the US.
實際上,有幾家公司在美國開始發展這項技術。
10:42.333–10:43.467
But how do they do it?
但他們是如何做到的呢?
10:43.467–10:45.467
So basically, you reflux the water.
基本上,就是將水進行回流。
10:45.467–10:48.200
There is an entire recycling technology on that.
這方面已有完整的回收技術。
10:48.420–10:51.780
The more concerning part of this is actually energy.
更令人擔憂的部分其實是能源。
10:52.900–10:55.833
Because of the usage of AI is exponentially growing,
由於AI的使用量呈指數級增長,
10:55.833–10:58.500
the energy usage is also growing exponentially
能源消耗也呈指數級增長,
10:58.500–11:00.860
and we are actually filling it.
我們實際上正在填補這個缺口。
11:01.200–11:05.640
So a lot of it is due to poor demand forecasting of power.
這其中很大一部分是因為電力需求預測不佳。
11:05.840–11:08.507
If you think about, look at the region as a whole,
如果你從整體區域的角度來看,
11:08.507–11:11.440
countries, I mean, Singapore is already doing that.
我指的是國家,新加坡已經在這樣做了。
11:11.440–11:13.173
The next data centre you build,
你接下來建造的資料中心,
11:13.173–11:15.040
you have to have a certain energy,
必須具備一定的能源條件,
11:15.040–11:16.707
that means the energy you use,
這意味著你所使用的能源中,
11:16.707–11:19.107
some parts of it must come from renewables.
其中一部分必須來自可再生能源。
11:19.107–11:21.507
Then as you go to say Malaysia, for example,
然後,當你考慮到例如馬來西亞時,
11:21.507–11:23.373
then the question is what kind of,
問題就變成了什麼樣的能源組合,
11:23.373–11:25.040
they definitely have energy,
他們肯定擁有能源,
11:25.040–11:26.507
they have petroleum, oil,
他們擁有石油、原油,
11:26.507–11:29.773
so they can hydrocarbon so you can power some part of it then
因此他們可以使用碳氫化合物來為部分設施供電,
11:29.773–11:31.440
the question is do they want to
問題是他們是否願意
11:31.440–11:32.973
rely on other mixes because
依賴其他能源組合,因為
11:32.973–11:36.820
the there is no way now that uh just by using a carbon they are able
目前沒有任何方法可以僅憑使用碳基能源就
11:36.820–11:38.565
to run that they also you also
能夠運行,而且這也會
11:38.565–11:42.420
bring about the increase of climate change as well and malaysia
也帶來氣候變遷的加劇,而馬來西亞
11:42.420–11:45.582
has the ambition to be that ai hub in southeast asia and
則有成為東南亞AI樞紐的抱負,
11:45.582–11:47.620
it really wants to achieve that so i
並且確實希望達成此目標,所以我
11:47.620–11:52.753
think the government needs to balance between addressing the people's concerns while also
我認為政府需要在回應民眾擔憂與採取務實步驟實現成為人工智慧樞紐的夢想之間取得平衡
11:52.753–11:57.090
taking realistic steps to make those dreams a reality of being an AI hub.
採取務實的步驟,將成為AI樞紐的夢想化為現實。
11:57.090–11:58.180
So the Malaysian
因此,馬來西亞
11:58.180–12:04.660
government, the Prime Minister announced that the young people will be able to access a three-month
政府,首相宣布年輕人將能獲得三個月的
12:04.660–12:07.580
free subscription to some of the AI models and
部分AI模型的訂閱服務,並
12:07.580–12:10.500
that is basically to allow them to explore the
這基本上是為了讓他們能夠探索
12:10.500–12:15.306
AI models, play with it and even to perhaps boost their creativity and
AI模型,進行體驗,甚至可能激發他們的創造力並
12:15.306–12:17.460
improve their digital skills.
提升他們的數位技能。
12:17.460–12:22.520
So the government is taking an active step into making the country embrace AI.
因此,政府正採取積極措施,讓國家擁抱AI。
12:22.740–12:26.980
But I guess we'll see how the government addresses the people's concerns.
但我猜我們會看到政府如何處理民眾的擔憂。
12:27.760–12:30.795
Will they keep building data centers for now with,
他們目前是否會繼續建設數據中心,
12:30.795–12:33.180
I guess, stricter regulations, Isaac?
我認為,艾薩克,這需要更嚴格的監管?
12:33.440–12:35.107
I think so, honestly, because
老實說,我認為會如此,因為
12:35.107–12:38.040
the people's concerns about data centers are valid.
民眾對數據中心的擔憂是合理的。
12:38.040–12:38.973
And for example,
舉例來說
12:38.973–12:44.240
you heard from Ether herself that she said that she doesn't want Malaysia to become like the US,
你聽過艾瑟本人說過,她表示不希望馬來西亞變成像美國那樣,
12:44.240–12:47.440
where the people actively push back against data centers
人們積極抵制數據中心的地方
12:47.440–12:50.561
because she's questioning why is data centers, AI,
因為她質疑數據中心、AI
12:50.561–12:52.460
relevant or even important?
為何相關甚至重要?
12:52.460–12:56.420
Because we've heard from the AI company CEOs themselves
因為我們聽過AI公司執行長本人說
12:56.420–12:59.353
that they say that this AI revolution is inevitable.
他們表示這場AI革命是不可避免的。
12:59.620–13:01.220
But she's questioning that.
但她對此提出質疑。
13:01.320–13:02.943
So I think we'll hear more and
所以我想我們會聽到越來越多
13:02.943–13:05.580
more people having similar concerns as Ether.
像艾瑟一樣有類似擔憂的人。
13:06.000–13:09.700
One interesting example of a country that's in this AI revolution
一個有趣的例子是,在這次AI革命中
13:09.700–13:12.560
that's not very much we would think about is China.
一個我們不太會想到的國家是中國。
13:12.560–13:15.820
so where China is pretty effective
中國相當有效的地方
13:15.820–13:17.520
is the energy infrastructure
在於能源基礎設施
13:17.520–13:18.653
so they have built up
他們已經建立
13:18.653–13:20.587
most of the energy infrastructure
大部分的能源基礎設施
13:20.587–13:21.587
they have built up
他們已經建立
13:21.587–13:22.187
most of the
大部分的
13:22.187–13:23.460
renewable energy mix
可再生能源組合
13:23.460–13:24.620
what they don't have
他們現在缺乏的是
13:24.620–13:25.487
is now the chips
現在是晶片
13:25.487–13:27.160
and the AI models are lagging
而AI模型落後
13:27.160–13:28.820
basically 3 to 6 months
基本上落後3到6個月
13:28.820–13:30.353
depending on who you talk to
這取決於你問的是誰
13:30.353–13:31.120
on that
關於這一點
13:31.120–13:32.820
but if you look at say
但如果你看看說
13:32.820–13:34.400
there was a survey done
有一項調查顯示
13:34.400–13:36.420
80% of Chinese
80%的中國人
13:36.420–13:38.320
are hopeful about the AI future
對AI未來充滿希望
13:38.320–13:40.600
and then if you do the same survey
然後如果你進行同樣的調查
13:40.600–13:42.080
in the US it's only 31%
在美國只有31%
13:42.080–13:45.080
So the question is, what's the big difference, right?
所以問題是,到底有什麼巨大的差異,對吧?
13:45.080–13:48.300
So I think a lot of it is in whether we're optimistic
我認為這很大程度上取決於我們是否樂觀
13:48.300–13:52.820
that the AI will bring about sort of a growth of a new economy.
認為 AI 將帶來一種新經濟體系的增長。
13:53.660–13:59.520
And there are some things that I think people take it to the logical extreme.
我認為有些人會將這一點推演到邏輯極端。
13:59.840–14:03.860
If AI was to do everything of this, then what should humans do, right?
如果 AI 能完成所有這些工作,那麼人類該做什麼呢,對吧?
14:04.180–14:07.920
I think the way I would probably want to moving ahead,
我認為我未來可能希望採取的方式是,
14:07.920–14:12.260
I think that there is actually human creativity is actually infinite.
我認為人類的創造力其實是無限的。
14:12.580–14:15.647
The demand for human creativity is actually infinite.
人類創造力的需求其實是無限的。
14:15.647–14:17.939
And I think for that, because of that,
我認為正因如此
14:17.939–14:20.760
even if AI can automate most of the things,
即使 AI 能自動化大多數事情,
14:21.100–14:23.033
there are still things for us to do.
仍然有我們要做的事情。
14:23.033–14:25.300
Yeah, really insightful conversation.
是的,非常有洞察力的對話。
14:25.300–14:27.000
Thank you so much for joining me, Bernad and Isaac.
非常感謝 Bernad 和 Isaac 來與我交流。

影片筆記:為何亞洲正大舉興建AI數據中心?馬來西亞成主要樞紐- BBC News 中文

<!-- VIDEONOTE_PUBLICATION_WARNINGS_BEGIN -->

⚠️ 可先使用,仍有待修正項目

本筆記已完成可閱讀的逐字稿與時間跳轉;下列為待處理問題,建議播放原片確認。

  • 待修正:逐字稿全文快速校讀未完成(已保留自動字幕/原始逐字稿並繼續生成筆記。;未完成的全文速讀項目可在精進模式回聽確認。)

<!-- VIDEONOTE_PUBLICATION_WARNINGS_END -->

一句話總結

馬來西亞憑藉鄰近新加坡的地理優勢及數據主權法規,成為東南亞 AI 數據中心樞紐,但引發當地居民對水資源消耗、電力需求及就業貢獻的強烈反對,同時全球對 AI 發展的態度在中西方呈現顯著差異。

核心重點

  • 馬來西亞成為樞紐的原因:
  • 新加坡禁令:新加坡 2019 年實施臨時禁令(moratorium),迫使開發商轉向鄰近的馬來西亞柔佛州。目前馬來西亞 85% 的數據中心位於柔佛州。
  • 數據主權法:2019 至 2023 年間,東南亞國家(如馬來西亞、印尼)實施數據主權法,要求數據在當地託管,迫使大型科技公司(Hyperscalers)在當地設立數據中心。
  • 地緣政治:2024 年後美國禁止中國獲取 NVIDIA 晶片,導致中國科技公司尋找其他算力來源,進一步推動外資進入馬來西亞。
  • 經濟與政府激勵:
  • IMF 預測馬來西亞當年 GDP 增長至少 4.7%。
  • 政府提供所得稅豁免(income tax exemptions)及投資補貼(investment allowances),目標成為 AI 革命先驅。
  • 馬來西亞總理宣布年輕人可免費獲得某些 AI 模型三個月訂閱,以提升數字技能。
  • 社會與環境爭議:
  • 居民反對:當地居民(如英語教師 Itha Malayana)發起請願,擔憂環境破壞、社區干擾及技能轉移不足。引用名言:「當最後一棵樹倒下,人們會意識到他們不能吃錢。」
  • 資源消耗:部分數據中心每日耗水 5000 萬公升(相當於 2500 萬人的日常飲水需求)。柔佛州政府已禁止一級(Tier one)和二級(Tier two)數據中心建設。
  • 能源效率:新加坡強制 PUE(電力使用效率)低於 1.25,馬來西亞約為 1.4。若實際使用量未達申報量 85%,將對運營商處罰。
  • 能源來源:馬來西亞擁有石油和碳氫化合物資源,但過度依賴碳能源會加劇氣候變化;面臨混合能源來源挑戰。
  • 就業與經濟價值爭議:
  • 建設期 vs 運營期:建設期創造建築工作,但運營期由大型科技公司接管,僅需極少數人員處理異常情況。
  • 技能轉移質疑:居民質疑 AI 用於日常決策(如午餐、關係問題)的必要性,認為數據中心並非真正的技能轉移或產業發展。
  • 數字經濟:數據中心真正目的是推動「數字經濟」,可能創造知識型工作或新 AI 相關工作,但目前處於過渡階段,擔憂「就業取代」大於「就業創造」。
  • 全球 AI 態度差異:
  • 中國:能源基礎設施完善,可再生能源組合已建立,但晶片和 AI 模型落後約 3 到 6 個月。調查顯示 80% 的中國人對 AI 未來持樂觀態度。
  • 美國:調查中僅有 31% 的人持樂觀態度。
  • 差異原因:在於是否樂觀於 AI 能帶來新經濟體系的增長。
  • 人類價值:即使 AI 自動化大部分工作,人類的創造力及其需求是無限的,人類仍有事可做。

詳細大綱

A. 馬來西亞成為 AI 數據中心樞紐的背景

  • 地理位置與新加坡因素:
  • 馬來西亞(特別是柔佛州)因鄰近新加坡,成為新加坡 2019 年實施「臨時禁令」(moratorium)後的替代地點。
  • 目前馬來西亞 85% 的數據中心位於柔佛州。
  • 法規驅動(數據主權):
  • 2019 至 2023 年間,東南亞國家(馬來西亞、印尼等)實施「數據主權法」(Data Sovereignty Laws),要求數據必須在當地託管。
  • 這迫使大型科技公司(Hyperscalers)在當地建立數據中心以開展業務。
  • 外部地緣政治因素:
  • 2024 年後,美國禁止中國獲取 NVIDIA 晶片,導致中國科技創始人尋找其他算力來源,進一步推動外資進入馬來西亞。

B. 經濟影響與政府激勵

  • 經濟增長:
  • 國際貨幣基金組織(IMF)預測馬來西亞當年 GDP 增長至少 4.7%。
  • 數十億美元投入馬來西亞經濟。
  • 政府激勵措施:
  • 提供所得稅豁免(income tax exemptions)。
  • 提供投資補貼(investment allowances)。
  • 目標是成為 AI 革命的先驅。
  • 年輕一代 AI 普及:馬來西亞總理宣布,年輕人可免費獲得某些 AI 模型的三個月訂閱,目的:讓年輕人探索 AI、發揮創意、提升數字技能。

C. 社會反彈與居民擔憂

  • 居民案例:
  • 26 歲英語教師 Itha Malayana 在社區發現數據中心提案後發起請願。
  • 引用名言:「當最後一棵樹倒下,人們會意識到他們不能吃錢。」
  • 擔憂點:環境影響、社區干擾、技能轉移不足。
  • 技能轉移質疑:
  • 居民認為數據中心僅需極少數人管理伺服器,並非真正的技能轉移或產業發展。
  • 質疑 AI 用於日常決策(如午餐、關係問題)的必要性。
  • 專家觀點與社會反彈:
  • 引用 Ether 的觀點:部分民眾質疑 AI 和數據中心對社會的相關性或重要性。
  • 擔憂馬來西亞可能像美國一樣,民眾積極推動反對數據中心。
  • 預計會有更多人產生類似 Ether 的擔憂。
  • 政府可能在更嚴格的法規下繼續建設數據中心,因為民眾的擔憂是合理的。

D. 專家觀點:就業與經濟價值

  • 就業結構變化:
  • 建設期:創造建築工作。
  • 運營期:由大型科技公司(Hyperscalers)接管,僅需極少數人員處理異常情況。
  • 數字經濟與就業創造:
  • 數據中心的真正目的是推動「數字經濟」(Digital Economy),可能創造知識型工作或新 AI 相關工作。
  • 目前處於過渡階段,擔憂「就業取代」大於「就業創造」。
  • 歷史技術變革通常最終創造更多工作,但短期內價值創造對本地經濟的貢獻尚不明確。
  • 數據主權與穩定性:
  • 即使數據來自全球,企業仍需投資當地以維持「可持續持有」(sustainable holding)。
  • 政府需平衡居民反對情緒(如抗議、破壞設施風險)與發展願景。

E. 環境挑戰與技術解決方案

  • 水資源消耗:
  • 部分數據中心每日耗水 5000 萬公升,相當於 2500 萬人的日常飲水需求。
  • 柔佛州政府已禁止一級(Tier one)和二級(Tier two)數據中心建設。
  • 能源效率與指標:
  • PUE 指標:電力使用效率。新加坡強制要求低於 1.25,馬來西亞約為 1.4。
  • 電力罰則:若實際使用量未達申報量的 85%,將對運營商進行處罰。
  • 技術與能源來源:
  • 冷卻技術:提及「StatePoint Liquid Cooling」,涉及水回流與回收技術。
  • 能源結構:馬來西亞擁有石油和碳氫化合物資源,但過度依賴碳能源會加劇氣候變化。
  • 可再生能源:新加坡要求部分能源來自可再生資源,馬來西亞面臨混合能源來源的挑戰。
  • 電力預測:AI 使用量指數級增長導致電力需求激增,存在電力需求預測不足的問題。

F. 全球 AI 發展現狀與態度調查

  • 中國現狀:
  • 優勢:能源基礎設施完善,可再生能源組合已建立。
  • 劣勢:晶片和 AI 模型落後約 3 到 6 個月(視來源而定)。
  • 民意:80% 的中國人對 AI 未來感到充滿希望。
  • 美國現狀:
  • 民意:僅 31% 的人對 AI 未來感到充滿希望。
  • 態度差異根源:是否相信 AI 能促進新經濟體系的增長。
  • 人類在 AI 時代的定位:
  • 針對「AI 自動化一切後人類價值何在」的討論。
  • 核心觀點:人類的創造力是無限的,對人類創造力的需求也是無限的。
  • 結論:即使 AI 自動化大部分工作,人類仍有其存在的空間和任務。

工具 / 模型 / 名詞整理

  • NVIDIA 晶片 (NVIDIA chips)
  • ChatGPT
  • IMF (International Monetary Fund / 國際貨幣基金組織)
  • PUE (Power Usage Effectiveness / 電力使用效率指標)
  • StatePoint Liquid Cooling (StatePoint 液體冷卻技術)
  • Tier one / Tier two (一級/二級數據中心分類)
  • Hyperscalers (大型科技公司/超規模計算提供商)
  • Data Sovereignty Laws (數據主權法)
  • moratorium (臨時禁令/暫停令)
  • Ether (人名/專有名詞,用於引用民眾觀點)
  • neocloud (非標準術語,逐字稿中提及 "whether it's a neocloud or hyperscaler")

操作流程整理

  • 數據中心建設流程:
  1. 法規驅動(數據主權法)或地緣政治因素(晶片禁令)導致需求。
  2. 大型科技公司(Hyperscalers)或新創公司(neocloud)評估地點(如馬來西亞柔佛州)。
  3. 申請政府激勵措施(所得稅豁免、投資補貼)。
  4. 建設階段(創造建築就業)。
  5. 運營階段(由大型科技公司接管,僅需極少數人員管理,依賴特定冷卻技術如 StatePoint Liquid Cooling)。
  6. 面臨當地居民反對(請願、抗議)及政府監管(PUE 指標、電力罰則、建設禁令)。

值得注意的限制或風險

  • 水資源風險:部分數據中心每日耗水量極大(5000 萬公升),可能影響當地居民飲水需求。
  • 電力供應風險:AI 使用量指數級增長導致電力需求激增,存在預測不足風險;過度依賴碳能源加劇氣候變化。
  • 社會穩定風險:當地居民反對情緒可能導致抗議或破壞設施,政府需平衡發展願景與居民反對。
  • 就業結構風險:運營期所需人員極少,可能導致「就業取代」大於「就業創造」,引發技能轉移不足的擔憂。
  • 法規監管風險:柔佛州政府已禁止一級和二級數據中心建設;若實際電力使用量未達申報量 85%,將面臨處罰。

逐字稿辨識疑點

  • Itha Malayana:逐字稿中提到的居民姓名,聽起來可能為音譯或特定人名,需查證是否為正確拼寫。
  • Aoife:在討論居民觀點時,講者提到 "what Aoife said",但前文提到的居民是 "Itha Malayana",此處人名不一致,疑點。
  • neocloud:逐字稿中提到 "whether it's a neocloud or hyperscaler","neocloud" 為非標準術語或聽寫錯誤,需查證。
  • 80 to 85%:講者先說 "85%",後又說 "about 80 to 85%",數據表述略有出入。
  • 50 million litres:逐字稿中提及 "50 million litres of water a day",此數值極大,需查證是否為特定大型數據中心的數據或聽寫錯誤(如是否為 50,000 或其他單位)。
  • 25 million people:對應 5000 萬公升水的飲用需求人數,此換算關係需查證。
  • 1.25 PUE / 1.4 PUE:新加坡與馬來西亞的 PUE 數值,需查證當前官方標準是否準確。
  • StatePoint Liquid Cooling:此名稱在業界較少見,可能為特定公司名稱或聽寫錯誤,需查證。
  • Ether:逐字稿中出現 "you heard from Ether herself",此處可能為人名或特定稱呼,聽起來較像 "AI" 或特定專有名詞的聽誤,但依規則保留原樣。
  • Bernad:結尾處感謝對象為 "Bernad",此名稱較不常見,可能為聽寫錯誤或特定人名。
  • 3 to 6 months:關於中國 AI 模型落後時間的描述,聽起來較短,可能為口誤或特定語境,依規則保留。
  • 80% / 31%:調查數據來源未明確標註,僅提及 "survey done"。

可延伸追問

  • 馬來西亞政府如何具體執行對一級和二級數據中心的禁令?是否有例外情況?
  • StatePoint Liquid Cooling 技術的具體運作原理及其在馬來西亞數據中心的實際應用案例為何?
  • 中國 AI 模型落後 3 到 6 個月是基於哪些具體指標或比較對象?
  • 馬來西亞總理提供的免費 AI 模型訂閱具體包含哪些服務?如何評估其對數字技能的實際提升效果?
  • 數據主權法在東南亞各國的具體執行細節有何差異?對跨國科技公司的合規成本影響為何?

生字列表

生字讀音類型中文
unprecedented/ʌnˈpres.ɪ.den.tɪd/前所未有的
emerged/ɪˈmɜːrdʒd/崛起;出現
aggressively/əˈɡres.ɪv.li/積極地;大刀闊斧地
pushback/ˈpʊʃ.bæk/反彈;抵制
justified/ˈdʒʌs.tɪ.faɪd/合理的;有根據的
slice/slaɪs/一份;一塊
proximity/prɒkˈsɪm.ə.ti/接近;鄰近
incentives/ɪnˈsen.tɪvz/誘因;激勵措施
prudent/ˈpruː.dənt/謹慎的;穩健的
spike/spaɪk/激增;尖峰
sovereignty/ˈsɒv.rɪn.ti/主權
hyperscalers/ˈhaɪ.pə.skeɪ.lərz/超大型雲端服務商
iteration/ˌɪt.əˈreɪ.ʃən/迭代;版本
displacement/dɪsˈpleɪs.mənt/取代;流失
pitchforks/ˈpɪtʃ.fɔːks/草叉
mandating/ˈmæn.deɪ.tɪŋ/強制規定
exponentially/ˌek.spəˈnen.ʃəl.li/指數級地
inevitable/ɪˈnev.ɪ.tə.bəl/不可避免的

生字解說

unprecedented /ʌnˈpres.ɪ.den.tɪd/

· C1

意思:前所未有的

解說:形容事物發生或出現的程度是歷史上從未有過的,強調其獨特性和規模。

影片原句
The AI boom means that data centers are being built at an unprecedented pace around the world
人工智慧熱潮意味著全球正以前所未有的速度建設數據中心
延伸例句
The company faced unprecedented challenges during the pandemic.
公司在疫情期間面臨前所未有的挑戰。

emerged /ɪˈmɜːrdʒd/

· B2

意思:崛起;出現

解說:指從隱藏或次要狀態變得顯著或重要,常用於描述新興趨勢或領導地位。

影片原句
Malaysia has emerged as an AI hub in Southeast Asia.
馬來西亞已崛起為東南亞的人工智慧樞紐。
延伸例句
She emerged as a leader in the tech industry.
她崛起為科技業的領袖。

aggressively /əˈɡres.ɪv.li/

· B2

意思:積極地;大刀闊斧地

解說:形容採取行動時充滿活力、決心強且毫不猶豫,常帶有擴張或競爭意味。

影片原句
The country has also been aggressively building data centers.
該國也積極建設數據中心
延伸例句
The marketing team is aggressively targeting new customers.
行銷團隊正積極鎖定新客戶。

pushback /ˈpʊʃ.bæk/

· C1

意思:反彈;抵制

解說:指對某項計劃或變化的反對意見或抵抗行為,通常來自受影響的群體。

影片原句
but there's some pushback too.
但也出現了一些反彈。
延伸例句
There was significant pushback from the local community.
當地社區出現了強烈的反彈。

justified /ˈdʒʌs.tɪ.faɪd/

· B2

意思:合理的;有根據的

解說:形容某種觀點、情緒或行動是有充分理由支持的,並非無端猜測。

影片原句
the big question is whether or not those concerns about data center are justified.
而最大的問題在於,人們對數據中心的擔憂是否合理。
延伸例句
His fears were justified when the project failed.
當專案失敗時,他的擔憂被證明是合理的。

slice /slaɪs/

· B2

意思:一份;一塊

解說:在此處為比喻用法,指從整體利益或市場中分得的一份份額。

影片原句
how is the pushback against them affecting Southeast Asia's effort to get a slice of the AI boom?
此外,這種反彈如何影響東南亞在人工智慧熱潮中分一杯羹的努力?
延伸例句
We need to secure a larger slice of the market.
我們需要確保獲得更大的市場份額。

proximity /prɒkˈsɪm.ə.ti/

· C1

意思:接近;鄰近

解說:指地理位置上的靠近,常作為優勢條件出現在商業或戰略討論中。

影片原句
Its proximity to Singapore makes it a perfect place to build data centers,
它靠近新加坡的地理位置,使其成為建立資料中心的理想地點,
延伸例句
The proximity to the airport is a major selling point.
靠近機場是一個主要的賣點。

incentives /ɪnˈsen.tɪvz/

· B2

意思:誘因;激勵措施

解說:指政府或機構提供的獎勵(如稅收減免、補貼),以鼓勵特定行為或投資。

影片原句
Furthermore, the government put out incentives
此外,政府推出了各種誘因
延伸例句
Tax incentives attracted many foreign companies.
稅收誘因吸引了許多外國公司。

prudent /ˈpruː.dənt/

· C1

意思:謹慎的;穩健的

解說:形容行事小心、考慮周全,特別是在處理風險或資源時。

影片原句
because I think the Singapore government is very prudent
因為我認為新加坡政府對於電力
延伸例句
It is prudent to save money for emergencies.
為緊急情況存錢是謹慎的做法。

spike /spaɪk/

· B2

意思:激增;尖峰

解說:指數量或強度的突然且劇烈的上升,常用於描述需求、價格或數據。

影片原句
They're mostly caused because of a spike in demand
這主要是由於運算需求激增所導致
延伸例句
There was a spike in electricity usage during the heatwave.
熱浪期間用電量出現激增。

sovereignty /ˈsɒv.rɪn.ti/

· C1

意思:主權

解說:在此語境下指國家對數據的控制權和管轄權,即數據必須存放在國內。

影片原句
countries started to have what we call data sovereignty laws.
國家開始實施我們所謂的「數據主權法」。
延伸例句
Data sovereignty is a key issue for multinational corporations.
數據主權是多國企業的一個關鍵問題。

hyperscalers /ˈhaɪ.pə.skeɪ.lərz/

· C1

意思:超大型雲端服務商

解說:指擁有龐大數據中心基礎設施的大型科技公司(如 AWS, Azure, Google Cloud)。

影片原句
prompts the hyperscalers to say, okay,
促使超大型雲端服務商表示,好吧,
延伸例句
Hyperscalers are driving the demand for server chips.
超大型雲端服務商正在推動伺服器晶片的需求。

iteration /ˌɪt.əˈreɪ.ʃən/

· C1

意思:迭代;版本

解說:指技術或產品的重複改進過程,每一輪改進稱為一次迭代。

影片原句
But I think we've seen different iterations of technological change.
但我認為我們已經看到了不同階段技術變革的迭代。
延伸例句
The software went through several iterations before launch.
該軟體在發布前經過了多次迭代。

displacement /dɪsˈpleɪs.mənt/

· C1

意思:取代;流失

解說:在此指因技術進步導致原有工作崗位消失或人員被取代的現象。

影片原句
the big fear is the job displacement is larger than the job creation.
因此,最大的擔憂是就業流失大於就業創造。
延伸例句
Automation causes job displacement in manufacturing.
自動化導致製造業的工作流失。

pitchforks /ˈpɪtʃ.fɔːks/

· B2

意思:草叉

解說:比喻用法,源自農民抗議的意象,指當地居民強烈反對或抗議的行為。

影片原句
the local communities holding pitchforks and say,
當地社區手持草叉高呼,
延伸例句
The locals are ready to grab their pitchforks against the new factory.
當地居民準備好拿起草叉反對新工廠的建設。

mandating /ˈmæn.deɪ.tɪŋ/

· C1

意思:強制規定

解說:指由政府或權威機構通過法規強制要求執行的行為。

影片原句
currently Singapore is mandating
目前新加坡強制規定
延伸例句
The city is mandating solar panels on new buildings.
該市強制要求新建築安裝太陽能板。

exponentially /ˌek.spəˈnen.ʃəl.li/

· C1

意思:指數級地

解說:形容增長速度極快,隨時間呈幾何級數增加,常用於描述科技趨勢。

影片原句
the energy usage is also growing exponentially
能源消耗也呈指數級增長
延伸例句
Data storage needs are growing exponentially.
數據存儲需求正在指數級增長。

inevitable /ɪˈnev.ɪ.tə.bəl/

· B2

意思:不可避免的

解說:指某事必然發生,無法阻止或改變。

影片原句
that they say that this AI revolution is inevitable.
他們表示這場AI革命是不可避免的。
延伸例句
Change is inevitable in the modern world.
在現代世界,改變是不可避免的。

句型解說(含實例)

means that [clause]

意思:意味著……

接續:Subject + means that + Subject + Verb

解說:用於解釋某事物所導致的結果或含義,後接完整子句說明具體情況。

影片原句
The AI boom means that data centers are being built at an unprecedented pace around the world
人工智慧熱潮意味著全球正以前所未有的速度建設數據中心
實例
  1. The new policy means that employees must work remotely.
    新政策意味著員工必須遠程辦公。
  2. His silence means that he agrees with the proposal.
    他的沉默意味著他同意該提案。

whether or not [clause]

意思:無論是否……;……是否

接續:Question/Statement + whether or not + Subject + Verb

解說:用於表達對某種情況是否發生的疑問或確認,強調兩種可能性。

影片原句
the big question is whether or not those concerns about data center are justified.
而最大的問題在於,人們對數據中心的擔憂是否合理。
實例
  1. We need to decide whether or not to invest in the project.
    我們需要決定是否投資該專案。
  2. It remains to be seen whether or not the plan will succeed.
    該計劃是否會成功仍有待觀察。

has to come from somewhere

意思:必須來自某個地方

接續:Subject + has to + Verb + from + somewhere/somewhere specific

解說:強調事物有其來源或根源,常用於解釋現象背後的實質基礎。

影片原句
Every application you use, whether it's with AI or without AI, has to come from somewhere.
你使用的每個應用程式,無論是否涉及人工智慧,都必須來自某個地方。
實例
  1. Good food has to come from fresh ingredients.
    美食必須來自新鮮食材。
  2. Success has to come from hard work.
    成功必須來自努力工作。

put out [noun]

意思:推出;發布

接續:Subject + put out + Object

解說:口語中常用來表示發布產品、政策或資訊,相當於 release 或 issue。

影片原句
Furthermore, the government put out incentives
此外,政府推出了各種誘因
實例
  1. The company put out a new version of the app.
    公司發布了應用程式的新版本。
  2. They put out a statement regarding the incident.
    他們就該事件發布了一份聲明。

stand out

意思:突出;顯眼

接續:Subject + stand out + (in terms of / as)

解說:形容某人或某事物在群體中特別出色或引人注目。

影片原句
Malaysia kind of stands out in terms of the number of data centers being built.
就建設數據中心的數量而言,馬來西亞顯得尤為突出。
實例
  1. Her talent really stands out in the class.
    她的才華在班上非常突出。
  2. This design stands out from the competition.
    這個設計在競爭中脫穎而出。

pour into [place]

意思:湧入;大量投入

接續:Subject (money/investments) + pour into + Location

解說:比喻資金或資源大量且快速地進入某個地區或領域。

影片原句
That all the investments started pouring into the country
所有投資開始湧入該國
實例
  1. Tourists pour into the city during the festival.
    遊客在節日期間湧入該城市。
  2. Venture capital is pouring into the startup sector.
    創投資金正湧入初創企業領域。

take [something] to the logical extreme

意思:將……推演到邏輯極端

接續:Subject + take + Object + to the logical extreme

解說:指將某個觀點或行為推導至其最極端、甚至荒謬的結論。

影片原句
And there are some things that I think people take it to the logical extreme.
我認為有些人會將這一點推演到邏輯極端。
實例
  1. He takes efficiency to the logical extreme by working 24/7.
    他通過全天候工作將效率推演到邏輯極端。
  2. Don't take convenience to the logical extreme at the cost of health.
    不要為了健康代價將便利性推演到邏輯極端。

take [action] to [purpose]

意思:採取……以達到……目的

接續:Subject + take + Action + to + Verb

解說:描述為了實現特定目標而採取的具體行動或步驟。

影片原句
So the government is taking an active step into making the country embrace AI.
因此,政府正採取積極措施,讓國家擁抱AI。
實例
  1. The company took steps to improve customer service.
    公司採取措施改善客戶服務。
  2. We need to take action to reduce pollution.
    我們需要採取行動減少污染。