Chinese AI Pioneer Questions OpenAI's Sustainability

中国人工智能领域先驱者质疑OpenAI的可持续性

Chinese entrepreneur and AI pioneer Kai-Fu Lee says he questions the sustainability of OpenAI, and discusses the impact of DeepSeek on global AI development. The former Google executive also gives his top AI picks on Bloomberg Television.

中国企业家、人工智能领域先驱李开复表示，他对OpenAI的可持续性存疑，并探讨了DeepSeek对全球人工智能发展的影响。这位前谷歌高管还在彭博电视台节目中分享了他心目中顶尖的人工智能项目。

I guess the race for supremacy stands between China and the United States right now?

Well, the clearly in both US and China, the pre-training of a giant model has consolidated and is consolidating. And it's become clear that open source will be the winner.

There's still many that will not concede right. OpenAI, Anthropic, who build their businesses believing they can build a better closed model than everyone else.

I think they got shocked when they saw a model as good. They would argue nearly as good. But either way, they're getting it for much, much lower cost.

我猜目前中美之间正在进行一场争夺霸主地位的竞赛，对吗？

嗯，很明显，在美国和中国，大型模型的预训练已经得到了巩固，并且还在持续巩固中。而且现在很清楚，开源会是最终的赢家。

不过仍然有很多公司不愿承认这一点。比如OpenAI和Anthropic，它们在开展业务时坚信自己能够打造出比其他所有人都更好的封闭模型。

我认为当他们看到一个同样优秀的模型时会感到震惊。他们可能会争辩说只是接近（那么好）。但不管怎样，对手能够以低得多得多的成本做到这一点。

If you think about OpenAI's costs of $7 billion of operating costs in 2024, DeepSeek probably operating with 2% the operating expense.

So the issue really isn't whose model is 1% better? I think they're all very good. But the issue is is OpenAI's model even sustainable, Right. I mean, you're spending $7 billion a year making a massive loss. And here you have a competitor coming in with an open source model that's for free.

And that company also is infinitely, infinitely lasting because this founder has enough money to fund it at the current level and has reduced the cost of computing by a factor of 5 to 10. So with that kind of formidable competitor, I think Sam Altman is probably not slee well.

Yeah, well, I do think I mean, obviously that DeepSeek would have really challenged some of the monetization strategies around AI, but at the same time, given that that is so commoditized now and in China as well, that open source approach, how many AI models do you think that the Chinese market can actually sustain?

Well, I think probably I'm guessing three: DeepSeek, Alibaba and Bytedance and they'll evolve over time.

想想看，OpenAI在2024年的运营成本达到了70亿美元，而DeepSeek的运营成本可能只有其2%。

所以问题其实并不是谁的模型要好1% ？我觉得它们都非常出色。但问题在于OpenAI的模型是否可持续，对吧。我的意思是，你每年要花费70亿美元，还面临着巨额亏损。而这时有一个竞争对手带着免费的开源模型进入市场。

而且那家公司还能长久地存续下去，因为它的创始人有足够的资金维持运营，并且计算成本降低了五到十倍。面对这样强大的竞争对手，我想山姆·奥尔特曼（Sam Altman）可能都睡不好觉了。

是的，我确实认为，显然DeepSeek会对围绕人工智能的一些盈利策略构成真正的挑战，但与此同时，考虑到现在无论是在全球还是在中国，开源方法已经如此普遍，你认为中国市场实际上能够支撑多少个人工智能模型呢？

嗯，我想大概是三个吧：DeepSeek、阿里巴巴和字节跳动，而且它们会随着时间不断发展。

Yeah, well there may be there will be many out there who aspire to build one. Right. And we'll see if there's a fourth player.

But even within these three, I think DeepSeek currently has has the momentum and that's why we're embracing that. But I also want to point out that in and I wouldn't say AI is commoditized, I would say the underlying Pioneer Model (Foundation model) that is trained with a huge amount of money. Is more commoditized.

It costs a lot and it's hard to to really compete with open source. But AI is not commoditized, right? It's like, you know, Windows kernel, Linux kernel that's kind of commoditized, but Windows is not commoditized. Microsoft makes a ton of money.

是的，可能会有很多公司渴望打造出自己的模型。没错。我们看看会不会出现第四个参与者。

但即便在这三家公司当中，我认为目前DeepSeek发展势头正猛，这就是我们看好它的原因。但我也想指出，我不会说人工智能已经商品化了，我会说那些花了大量资金训练的底层先驱模型（基础模型）更具有商品化的特征。

这花费巨大，而且很难真正与开源模型竞争。但人工智能本身并没有商品化，对吧？就好比，你知道的，Windows内核和Linux内核有点商品化的性质，但Windows系统本身并没有商品化。微软赚得盆满钵满。

以下是外国网友的评论：

pruthirajpradhan1295
Money and talent will lead to success.When western media talks about chinese ev,solar domination they only talk about subsidies but never about the massive talent coming out of china.

金钱与人才铸就成功。西方媒体谈及中国在电动汽车和太阳能领域的主导地位时，只提补贴之事，却对中国大量涌现的人才绝口不提。

Repeal209
As if EV and solar required talent

好像电动汽车和太阳能行业压根不需要人才似的。

oceanwave4502
Am I the only one noticing that both USA and China have their advantage? (USA: capital (funding), China: cheap talent, some capital). But EU has neither! (Mistral (France) is valued only at 6 billions! and that's the BEST EU startup in LLM)

难道只有我注意到，美国和中国各有优势吗？（美国：资金雄厚；中国：人才成本低，也有一定资金）。可欧盟却两者皆无！（法国的米斯特拉尔公司估值仅60亿美元！而这已是欧盟大语言模型领域最出色的初创企业了）。

vlhc4642
They also don't talk about how come China has infinite money without inflation lol

他们也从不探讨，中国为何能有源源不断的资金投入，却不见通货膨胀，真是可笑。

arunangshudebnath7139
Low talent density in EU. Distinct lack of multidisciplinary thinkers and doers (used to be the strength, if you think about it). Let's not mince words. Regulation is a barrier..but it comes much later in the process.

欧盟人才密度低。明显缺乏跨学科的思考者与实干家（想想看，这曾是欧盟的优势所在）。我们打开天窗说亮话，监管的确是个障碍，但这是发展进程后期才出现的问题。

zuriyel5368
American industries receive more in subsidies

美国产业获得的补贴更多。

ObservingBeauty
Fascinating and true in my view. I wonder why all the "greeners" dont praise the immense save of energy that using deepseek presents

在我看来，这既有趣又真实。我纳闷儿，为何所有的“环保主义者”都不称赞使用DeepSeek所带来的惊人的能源节省呢。

IvarDaigon
they dont need to because not only is it more efficient but its alway way cheaper to use open source models which means that anybody with half a brain is looking at how to incorporate them into their workflows.
its like asking why didnt environmentalists praise flat screen monitors when they came out versus the old chunky CRTs that were chunky and used way more energy.. but they really didn't need to because those things sold themselves.

开源模型不仅效率更高，而且成本总是低得多。

但凡有点脑子的人，都在琢磨如何将其融入自己的工作流程。这就好比，人们会问，平板显示器问世时，环保主义者为何不盛赞它，而非要跟又笨又重、耗能极高的老式阴极射线管（CRT）显示器相比呢。

但他们真没必要称赞，因为平板显示器靠自身优势就足以畅销。

paulbo9033
As a "greener", I do praise this. And look forward to using it.
I'm glad I could set the record straight for you.

身为一名“环保主义者”，我确实对其赞赏有加。而且我期待着使用它。

ObservingBeauty
Clear. Same. Our planet surely benefit from Deepseek innovative breakthrough.

明白。我也有同感。我们的地球肯定会从DeepSeek的创新突破中获益匪浅。

jx.s1771
chinese officials are evaluated based on their achievements, that is, economic development, while American government officials are evaluated based on votes, which is probably the essential difference between the two models.

中国官员的考核基于政绩，也就是经济发展情况。而美国官员的考核基于选票，这或许就是两种模式的本质区别。

KhunYasoboyYaso-i5c
And india base on boasting and bragging.

印度就只会自吹自擂。

skierpage
You're confusing politicians with government officials. A lot of American government is fairly well-run, despite Republicans hating the idea of government making life better for all citizens.
Local government in both countries is subject to corruption in handing out contracts, and in addition, Chinese local governments make a lot of bad investment and property decisions.

你把政客和政府官员混为一谈了。尽管共和党人讨厌政府让全体公民生活更美好的理念，但美国很多政府部门的运作相当不错。中美两国的地方政府在签订合同方面都存在腐败问题，此外，东大的地方政府也做过不少糟糕的投资和房地产决策。

baxterslabo
The USA will soon learn that they are not the only country in the world with smart and talented people. LOL. Shout out to the Chinese!!

美国很快就会明白，世界上并非只有他们国家拥有聪慧且有才华的人。哈哈。为中国人点赞！

OakhillSailor
Obviously they are, especially California to be exact. They innovate the rest follow.

显然是这样，确切地说尤其是加利福尼亚州。他们搞创新，别人跟着学。

Hopper-n5f
Any country with money can hire talents

只要有钱，任何国家都能招揽人才。

Mk2kRaven
China is 10 years ahead of the US in terms of city infrastructure.

在城市基础设施方面，中国领先美国十年。

hydrohasspoken6227
@Hopper-n5f i fear it is not that simple.
Pay talents well and invite them to England.
Pay the same people well and invite them to Japan.
I bet my House many would choose England. Japanese is a linguistic minefield and criminally tough to master.

我觉得事情没那么简单。给人才优厚待遇，邀请他们去英国；同样给优厚待遇，邀请他们去日本。我敢打包票，很多人会选英国。日语简直就是语言雷区，难学得出奇。

Hopper-n5f
@Mk2kRaven Lol, do you not know the real estate is collapsing in tofu China?

哈哈，你难道不知道中国的所谓“豆腐渣”房地产正在崩溃吗？

Hopper-n5f
@hydrohasspoken6227 Lol, Japan is one of the leading countries in robotics. Of course they have talents

哈哈，日本可是机器人领域的领先国家之一。他们当然有人才。

deeplife9654
Open AI costs $5 billion dollars a year to keep running. It is going to lose eventually

OpenAI每年运营成本高达50亿美元。最终必将失败。

strongchallenger2269
Sam: That's why I ask the congress to ban it! Permanently!

这就是我请求国会永久封禁它的原因！

chantow8657
@deeplife9654 because Sam and gang took a huge chunk of it, that’s why it costs so much

因为山姆和他那帮人拿走了很大一部分钱，所以成本才这么高。

oceanwave4502
When OpenAI collapses, does it mean the wealth transfer is complete? (from investors around the world to pockets of US CEO, US AI developer as well as US government (tax)).

当OpenAI倒闭时，是否意味着财富转移完成了呢？（从全球投资者转移到美国企业首席执行官、美国人工智能开发者以及美国政府税收的口袋里）。

augm3nt384
Open Source AI will win. Power requirements and lack of innovation will kill Closed AI companies.

开源人工智能会胜出。电力需求和缺乏创新，会拖垮封闭的人工智能公司。

BillyP13
Apple is still around!

苹果公司还好好地存在着呢！

cwaddle
No way, there is not enough capital for model training. China was nice to release it open source but if the next big breakthrough comes closed door, that will be the end of it

不可能，根本没有足够资金用于模型训练。中国将其开源是善举，但如果下一个重大突破来自封闭模式，那就完了。

zyansheep
@BillyP13 apple is not an ai company tho

可苹果公司又不是人工智能公司。

anatolwegner9096
They called their model 01 openly mocking openAIs o1

他们把自己的模型命名为“01”，这明显是在公然嘲讽OpenAI的“o1”。

luccalus
I like Kai Fu's frank talk. Right or wrong only the future will show, but he is putting out his opinion.

我欣赏李开复的坦率言论。对错与否，只有未来才能证明，但他敢于表达自己的观点。

4mb127
"Probably operated at 2% expense." I doubt that very much. How is this measured? Is this based entirely on the word of Deepseek?

“运营成本可能只有对手的2%。”

我对此深表怀疑。这是如何测算出来的呢？

难道仅仅是基于DeepSeek的一面之词？

wote100-bx3ct
The Chinese value action over words. Those who talk too much often know the least. Since DeepSeek’s launch, its CEO has avoided media attention. Lee lacks expertise in cutting-edge AI advancements—his focus is on quick profits, not innovation. In fact, he never even mentioned DeepSeek until it became a sensation

中国人重行胜于重言。夸夸其谈者往往所知甚少。自DeepSeek推出以来，其首席执行官一直避开媒体关注。李开复在人工智能前沿进展方面缺乏专业知识——他关注的是快速获利，而非创新。事实上，在DeepSeek引起轰动之前，他从未提及过它。

harriemeeuwis978
He didn't mention Manus, that was announced this week from China as better than Deepseek.

他没提到“Manus”，本周中国宣布的这个项目，据说比DeepSeek更出色。

ymhktravel
It's still in the testing phase I guess. Only tokens are given.

我猜它还处于测试阶段。目前只发放了代币。

frankfighton
Manus is not the same thing, it is an application built on top of an anthropic model, very useful but not comparable with DeepSeek

“Manus”可不是一回事，它是基于一种拟人模型构建的应用程序，非常实用，但无法与DeepSeek相提并论。

avidbosak7503
Wow, Mr. Lee is incredibly articulate, honest, open, and insightful. Haven't seen any leaders like this in US. US leaders obsessed with posturing, politics, and, of course, fear of Chinese competition.

哇，李开复先生口才绝佳、坦诚直率、思想开放且富有洞察力。美国可没有这样的领导人。美国领导人热衷于装腔作势、玩弄政治，当然，还惧怕中国的竞争。

akka2011hk
I just wonder why Li doesn't think Tencent and Hauwei will be amongst the winners. In terms of users Tencent has more users than both Alibaba and Byte dance. Hauwei has access to more properitary data as it is enmesh in most of China's industry and consumer base systems. Has he also consider that Deepseek may build its middleware one day and take the carpet right under his feet?

我就想知道，为什么李开复不认为腾讯和华为会成为赢家之一呢。从用户数量来看，腾讯的用户比阿里巴巴和字节跳动都多。华为融入了中国大多数工业和消费者基础系统，能获取更多专有数据。他有没有想过，DeepSeek有朝一日可能会开发自己的中间件，然后抢了他的饭碗呢？

umanthmurthy1642
He’s mostly correct. Sam Altman is a leech these days who just wants to make money without any proper strategy.

他说得大体没错。如今的山姆·奥尔特曼就像个吸血鬼，只想赚钱，却毫无恰当的策略。

mimihappier1095
Once again, Chinese are focusing on usefulness. They are building handy bots/agents/operators in varieties of fields such as Finance, EV industry, medical industry etc.

中国人再次注重实用性。他们在金融、电动汽车、医疗等各个领域打造实用的机器人、智能体或操作程序。

Sonder2030
More money is more better in life but especially in AI. Deep seek can’t serve their model, reasoning models require a lot of compute to run. No one mentions that which I should be surprised because it doesn’t fit the narrative.

生活中钱越多越好，在人工智能领域更是如此。DeepSeek可能无法支撑其模型运行，推理模型的运行需要大量的计算能力。没人提到这一点，我很惊讶，因为这不符合他们的说法。

yaoliang1580
China's phenomenal progress in Ai and other spheres of technologies pose a serious challenge to those American big tech oligarchs who want to have complete monopoly of Ai and all key technologies n thus able to make exorbitant profits in a world that's completely without any competition.
The American open Ai is a misnomer and should be renamed closed Ai

中国在人工智能及其他技术领域取得的惊人进展，对那些企图完全垄断人工智能及所有关键技术，从而在世界中获取暴利的美国大型科技寡头构成了严峻挑战。美国的“OpenAI”这个名字名不副实，应该改名为“封闭AI”。

2Cannonball
I think every country should invest in their own model. Just look at how unreliable and foolish it would be to become dependent on the US

我认为每个国家都应该投资自己的模型。想想看，依赖美国是多么不可靠、多么愚蠢的行为。

zack_120
The inventor of AI concept and technology can charge DS thro licensing the use of AI which they are good at but bad at inventing.

人工智能概念和技术的发明者，可以通过授权使用人工智能向DeepSeek收费，毕竟他们擅长的是收费，而非发明创造。

m.preacher2829
openai did it from strach and you just do it based on their model for granted and say they are not sustainable and stupid
no wonder why China cannot create something innovation

OpenAI是从零开始做的，而你们理所当然地基于他们的模型来做，还说他们不可持续、很愚蠢。难怪中国搞不出创新成果。

gengron6297
China has the world's largest number of science and engineering talents, with 12 million graduates last year. Competition for talent is fierce, and when it comes to immigration, the number of Chinese immigrants is even smaller than India, don't you have a problem with your immigration policy? China is not an english-speaking country.

中国拥有世界上数量最多的理工科人才，去年有1200万毕业生。人才竞争十分激烈，而且说到移民，中国的移民数量甚至比印度还少，难道你们的移民政策没有问题吗？中国又不是英语国家。

seancai1049
Open AI will be bankrupt soon, it's losing talent at a alarming rate, Microsoft is getting cold feet and to continue to spend 10 billion a year it need to keep rising it's valuation the latest round have it value at 380 Billion, what company that lose 5-7 billion a year is worth 380 B? when it reach that valuation that mean the next company coming in have to invest Open AI at that valuation just to keep the lights on, after soft bank who is even that stupid to keep investing in Open AI?

OpenAI很快就会破产，它正以惊人的速度流失人才，微软也开始打退堂鼓了。要想每年继续投入100亿美元，它就得不断提高估值，最新一轮估值已经达到3800亿美元。哪家每年亏损50到70亿美元的公司能值3800亿美元呢？当它达到这个估值时，意味着下一家投资OpenAI的公司必须按这个估值来投资，才能维持其运营。软银之后，还有谁会傻到继续给OpenAI投资呢？

继续阅读：人工智能

此文由 三泰虎 编辑，未经允许不得转载！：首页 > 资讯 » 美媒：中国人工智能领域先驱者，质疑OpenAI的可持续性