2025年1月20日，当西方科技巨头齐聚特朗普的总统就职典礼时，一家名不见经传的中国人工智慧（AI）公司——DeepSeek（深度求索)，悄然发布了其R1模型的人工智能开源版本，并附上技术报告及不受限制的商业使用许可证。
这不仅仅是普通的AI应用程式。其在多项核心指标上可以OpenAI的ChatGPT匹敌，甚至超越之。更重要的是，它对普通用户和企业用户而言完全免费，且可在本地部署。
所谓“免费开源”是指任何人都可以免费使用、修改和分发的软体，这意味著原始程式码可以公开使用，并且允许任何人更改，一般需要在允许此类修改和共享的许可证下进行；从本质上讲，这是一种程式码公开给任何人免费使用的软体。
“斯普特尼克”危机
DeepSeek的发布在AI技术领域引发了一场风暴，美国虽然是目前的AI霸主，但DeepSeek的诞生让市场大为震撼。DeepSeek立即成为美国及其他国家下载量最高的免费应用程式。
据《彭博社》报导，DeepSeek的AI助手应用程式迅速登上140个市场的下载排行榜首位，其中印度用户占比最大。1月26日，该应用在苹果App Store排名第一，并稳居全球榜首。此外，它也在1月28日夺得谷歌Play商店美国区的首位。短短18天内，下载量达1600万次，几乎是OpenAI ChatGPT当初推出时900万次的两倍。
DeepSeek的低成本、高性能让AI消费者（包括全球研究人员、大小企业、专业人士和网民）纷纷响应，但与此同时，其也冲击了全球股市，让投资者重新审视AI技术及相关股票的价值。
1月27日，美国股市就重估了在美国牛市中扮演重要角色的人工智能和晶片股的价值。那斯达克100指数下跌3%，标准普尔500指数下跌1.5%。其中一家备受关注的晶片制造商就创下2020年3月以来最大跌幅。美国人工智能热潮的典型代表英伟达，其股价就下跌17%。这是有史以来个股最大的市值损失，也是投资者对DeepSeek的开发和营运成本远低于Meta、Alphabet等美国人工智能公司即时反应，这些美国人工智能公司花在英伟达和其他晶片升级上费用可高达数十亿美元。
AI趋势数据
●全球77%的设备具备某种形式的AI功能
●90%的企业将AI视为一种竞争优势
●2030年，AI预计将为全球经济贡献15.7兆美元
●2025年，AI可能取代8500万个工作岗位，但同时创造9700万个新职位，增加1200万个新工作机会
●63%的企业 计划在未来三年内采用AI技术
●AI市场预计每年增长至少120%
●2024年，全球AI市场将同比增长33%
●88%的非用户不清楚生成式人工智能将如何影响他们的生活。
●只有1/3的消费者认为他们正在使用人工智能平台，而实际使用率高达77%。
（来源：https://www.nu.edu/blog/ai-statistics-trends/）
DeepSeek：中国AI新变数
DeepSeek的未来发展将不会是由美国和盟国的消费者或人工智能市场来决定。反之这将成为美国试图遏制中国崛起的新战场。
值得注意的是，就在特朗普就职典礼的次日，美国股市大跌之前，特朗普政府宣布了一项由OpenAI、软银和甲骨文（Oracle）共同成立的“星门”（Stargate）计划，该项目将投资高达5000亿美元于美国人工智能基础设施，如数据中心和服务器，确保美国在人工智能领域的领先地位。
特朗普称这是“史上最大规模的AI基础设施计划”，将确保“科技未来仍然掌握在美国手中”。
在美国股市暴跌后，特朗普在迈阿密的一场共和党国会议员闭门会议上表示，中国的人工智慧可以刺激美国公司创新。
“希望中国DeepSeek出现能让我们的产业清醒过来，让我们全力以赴竞争并取胜。”
“这可能是件好事，因为与其花费数十亿美元，我们或许可以以更少的成本，取得同样的成果。”
然而，特朗普是否会继续维持这种开放态度，仍然值得观察，因为美国内部的反华势力、亲台湾游说团体及利益相关企业，如台湾晶片制造商，可能会强力推动美国封锁DeepSeek。
DeepSeek面临的挑战与抹黑
对DeepSeek的第一枪已经射出。一段广为流传的社交媒体短片就指DeepSeek员工和同事在农历新年期间应对了一场长达83小时的大规模网络攻击，攻击者来自一个技术高度成熟、身份不明的美国黑客组织，使用了以下手段：
●通过“钓鱼”邮件试图盗取员工帐号
●暴力破解尝试入侵防火墙
●恶意数据注入包括勒索软体和间谍软体
虽然DeepSeek成功抵御了第一波的攻击，但接下来更艰难的是在美国的意识形态和经济利益的新结合下，他们 不只要确保DeepSeek被击退，更要被击倒。
这可从美国和西方媒体开始针对DeepSeek进行各种抹黑攻击印证出来，这些对DeepSeek指责有：
● 隐瞒实际研发成本
● 涉嫌窃取或侵犯知识财权
● 使用被禁AI技术与走私晶片
● 政治偏见，受中国政府控制
DeepSeek的捍卫者指出，这些指控几乎没有任何证据。对于，目前人工智能市场领导者OpenAI提出的主要指控，DeepSeek的“资料蒸馏”人工智能训练技术，涉及知识财权侵权或盗窃，这些指控突显了其虚伪一面，特别是OpenAI目前正面临十几起版权诉讼，指控其使用受版权保护的作品来训练其ChatGPT使用的大语言模型。
DeepSeek必须应对最具威胁的指控，指其对美国的国家安全和世界领导地位构成威胁。炒作这威胁论的包括美国国会领袖、大力投资当前人工智能技术的企业家、媒体上的“专家”，他们利用台湾、新疆、西藏等课题来声称DeepSeek“对言论自由极其危险”，因为这将让中国控制政治述论，以及反华阵营的言论。
在DeepSeek方面，我们看到了来自一些关键科技领袖的意外声援。人工智能晶片巨头英伟达首席执行员黄仁勋将DeepSeek的R1模式描述为“卓越的AI进步”，尽管他的公司及个人财富是矽谷科技股暴跌中最大输家。微软首席执行员纳德拉称赞DeepSeek为“真正的创新”；OpenAI的首席执行员奥尔特曼则表示DeepSeek“显然是一个出色的模型”；苹果首席执行员库克则表示“提高效率的创新是件好事”。
DeepSeek及其他中国新兴AI应用，如阿里巴巴的Qwen2.5v-VL所带来的最重要价值，是让AI的发展不再受限于任何单一国家或企业的掌控。提供免费开源AI，将是让所有拥有手机的人都能公平获取资讯与技能的关键。
这种潜力远远超越了美国批评者对中国AI的关注——无论是政治、经济还是军事影响。如果美国及其盟友选择限制或封禁DeepSeek，或像对待华为和TikTok那样将其视为新冷战的战场，那么最终受害的将是全球其他国家。
林德宜《美国能阻挡中国AI新星DeepSeek吗？》原文：Can the US Hold Back DeepSeek – China’s AI Superstar and Game Changer
On 20 January 2025, a day when the tech titans of the West were in prominent attendance at Trump’s presidential inauguration, a little known China artificial intelligence (AI) company DeepSeek released an open source version of its R1 model, a technical report as well as licencing for unrestricted commercial use.
This was no ordinary AI app. It matched and surpassed the leading AI app, OpenAI’s ChatGPT, on several key benchmarks. Perhaps most important to ordinary as well as business users of AI apps, it was free and locally deployable.
“Free open source” refers to software that is available for anyone to use, modify, and distribute without cost, meaning the source code is publicly accessible and can be changed by anyone, typically under a license that allows for such modifications and sharing; essentially, it’s software where the code is openly available and can be used freely by anyone.
Sputnik Moment Impact of DeepSeek
The response was as if a thunderbolt had been unleashed on the AI tech world in which the US is the dominant player and the world’s leader. The app immediately became the most downloaded free app in the US and other countries.
According to Bloomberg, DeepSeek’s artificial intelligence (AI) assistant topped the list of most downloaded mobile apps across 140 markets, with India accounting for the largest percentage of new users at the end of January. It had risen to the No 1 spot on Apple Inc’s App Store on Jan 26, and has held that position globally since. The app has also held the top spot on Alphabet Inc’s Android Play Store in the US since Jan 28. With 16 million downloads over its first 18 days, the app almost doubled the nine million notched by OpenAI’s ChatGPT when it was first released.
The positive response of AI consumers – researchers, small and big time businesses, professionals, and netizens all over the world – stands in contrast with global stock markets where the Chinese low budget AI is causing investors to question everything that they know about AI and related stocks and their valuation.
On 27 January, the US stock market recomputed the valuation on AI and chip stocks that have been a key player in the US bull market run. The Nasdaq 100 sank 3% while the S&P 500 dropped 1.5%. A closely watched gauge of chipmakers plunged the most since March 2020. Nvidia, the poster child of America’s AI frenzy, sank 17%. This was the biggest market-cap loss for a single stock ever as investors chewed over the revelation that DeepSeek was developed and operates at a fraction of the billions of dollars that US AI companies such as Meta, Alphabet and others have spent on Nvidia and other chips to upgrade their hardware.
Fast Facts About Artificial Intelligence
● 77% of devices being used have some form of AI.
● 9 out of 10 organizations support AI for a competitive advantage.
● AI is projected to contribute $15.7 trillion to the global economy by 2030.
● By 2025, AI might eliminate 85 million jobs but create 97 million new ones, resulting in a net gain of 12 million jobs.
● 63% of organizations intend to adopt AI globally within the next three years.
● AI market size is expected to grow by at least 120% year-over-year.
● In 2024, the global AI market is projected to grow 33% year over year.
● 88% of non-users are unclear how generative AI will impact their life.
● Only a third of consumers think they are using AI platforms, while actual usage is 77%.
Source: https://www.nu.edu/blog/ai-statistics-trends/
DeepSeek: Another China Game Changer
What comes next for DeepSeek may not be up to consumers or the market for AI based in the US and ally countries to decide. It may be the latest battlefield that the US sees a need to fight to stymie the rise of China as a world power.
It is significant that a day after his inauguration and before the US market meltdown, President Trump announced a joint venture between OpenAI, SoftBank and Oracle to create up to US$500 billion in computing infrastructure to power artificial intelligence. The venture, called Stargate, was to invest in US data centers and servers throughout the country.
According to Trump this was “the largest AI infrastructure project by far in history” and would keep “the future of technology” in the US.
Following the rout in Wall Street, Trump told a Republican congressional retreat in Miami that he thought the Chinese AI could act as a spur for the US companies to innovate.
“Hopefully, the release of DeepSeek AI from a Chinese company should be a wake-up call for our industries that we need to be laser-focused on competing to win”.
“I would say that could be a positive. So instead of spending billions and billions, you’ll spend less, and you’ll come up with, hopefully, the same solution.”
Despite this initial positive response, Trump may change his mind on welcoming Deepseek.
This is if the anti China and pro Taiwan lobby and bandwagon of business interests, including those invested in Taiwanese chip companies, get their way.
The opening shots of a likely sustained battle that DeepSeek has to fight has already been launched. A widely circulated social media video describes an 83 hour digital war fought by DeepSeek staff and associates during the Chinese New Year period to deal with a large scale cyber attack emanating from an unknown but highly sophisticated adversary based in the US using
● phishing emails sent to employees to steal credentials
● brute force attempts to crack passwords and breach firewalls.
● Injection of malicious data including ransomware and spyware
This first threat to DeepSeek was repulsed. What will be more difficult is the emerging alignment of ideological and financial interests that have emerged to ensure that DeepSeek is not only pushed back but brought down.
This can be seen in the numerous western media reports accusing DeepSeek of
● misleading about its true development cost
● intellectual property infringement or theft
● operating on restricted AI and smuggled chips
● political bias and control by China
Defenders of DeepSeek point out that the allegations have little or no substance. Also that the main charge levelled by current AI market leader, OpenAI, that DeepSeek has engaged in intellectual property infringement or theft through distillation of data smacks of hypocrisy and worse, especially since OpenAI is currently facing more than a dozen copyright lawsuits alleging use of copyright-protected works to train the large-language models used by its ChatGPT.
The most formidable of the criticisms DeepSeek has to deal with is that it is a threat to US national security and world leadership. Those beating the drums of war against DeepSeek on this include Congressional leaders, entrepreneurs heavily invested in the current AI technology, ‘experts’ fronted in the media, who use select loaded questions on Taiwan, Xinjiang, Tibet, etc. to claim that DeepSeek is “incredibly dangerous for free speech” as it allows China to control political narratives, and others in the anti China camp.
DeepSeek Defenders
On DeepSeek’s side, we have seen unexpected support from some key tech leaders. Nvidia ‘s CEO, Jensen Huang, described DeepSeek’s R1 as “an excellent AI advancement”, although his company and his own personal worth were the biggest losers from the Silicon Valley tech rout. Microsoft’s Satya Nadella praised it for its “real innovations” ; OpenAI’s Sam Altman described it as “clearly a great model”; whilst Apple CEO Tim Cook said “innovation that drives efficiency is a good thing”.
The most important benefit that DeepSeek and other new Chinese AI apps such as Alibaba’s Qwen2.5v-VL are providing is the democratization of AI beyond the control of any single country or company. Providing free open source AI will be key to levelling the playing field in access to information and skills for everyone with a handphone.
This potential is much more than the politics, economics or military implications that US critics of China’s AI are focusing on. By restricting or banning DeepSeek or in dealing with it as another new cold war front as with Huawei and TikTok previously, the US and its allies want the rest of the world to end as the losers.
