AI journey of China is a chapter in human progress
[Photo/VCG]
In recent years, artificial intelligence has emerged as a transformative force, reshaping industries and redefining the global technological landscape. Widely regarded as a cornerstone of the Fourth Industrial Revolution, AI is driving breakthroughs in fields such as healthcare, finance, manufacturing and transportation.
At the forefront of this revolution are some United States-based tech giants such as Nvidia, Google and Intel, whose sustained investments in research and development have so far solidified their leadership in the field.The optimism surrounding AI's potential has not only fueled technological advancements but also driven a significant increase in the stock prices of relevant companies, galvanizing investors worldwide.
Having said that, the global AI race is not a one-sided game. While the US has long been a dominant player, China has rapidly emerged as a formidable competitor, demonstrating remarkable resilience and innovation in the face of significant challenges. Chief among these challenges is the US' technology sanctions, which have limited China's access to advanced semiconductor chips.Everyone knows these high-performance chips are critical for training AI models. Many believed that such restrictions could severely hinder China's ambitions in this regard. Yet, China's talent has defied expectations, leveraging independent research, creative engineering and strategic investment to sustain its momentum in AI development.
Faced with a scarcity in advanced computing power because of restrictions on high-performance chips, Chinese engineers have adopted innovative approaches to optimize AI workflows.
Techniques such as model pruning, quantization and knowledge distillation have been employed to reduce the size and complexity of AI models while maintaining high levels of accuracy, enabling Chinese AI to operate efficiently on less-advanced hardware, effectively bypassing the need for top-tier chips.
For instance, DeepSeek has been designed to perform exceptionally well on Nvidia's H800 chips, which are less powerful than the topnotch H100 chips used by large language models such as ChatGPT.This achievement underscores the ingenuity of Chinese researchers and their ability to adapt to constrained environments.
Another critical aspect of China's AI strategy has been the development of domestic hardware alternatives. Companies such as Huawei have made significant strides in designing and producing AI-specific chips, such as the Ascend series. While these chips may not yet match the performance of their international counterparts, they are steadily narrowing the gap and providing a viable foundation for China's AI ambitions.Substantial investments in cloud computing and distributed computing infrastructure, which allow AI models to scale using networks of lower-performance processors, are also essential.
China's rise as an AI powerhouse is exemplified by its growing portfolio of cutting-edge AI models and technologies. Among these, Deep-Seek has garnered significant attention for its ability to push the boundaries of AI capabilities despite hardware limitations. Alongside other domestic AI models, such as Doubao, Kimi, Wenxin Yiyan and Tongyi Qianwen, they represent China's determination to forge a unique developmental path in the face of adversity.
The success of DeepSeek and other Chinese AI models highlights the importance of optimizing both hardware and software to achieve breakthroughs in AI. Focusing on algorithmic efficiency and leveraging domestic hardware solutions demonstrate that innovation can thrive even under significant constraints.
DeepSeek stands out as a prime example of China's innovative capacity in the AI domain. Built on advanced machine learning algorithms and extensive datasets, it excels in natural language processing, image recognition and predictive analytics. Since the release of its third iteration, DeepSeek has gained recognition for its exceptional capabilities, positioning itself as a strong contender against OpenAI's ChatGPT.
What sets DeepSeek apart is its ability to deliver high performance despite operating on less advanced hardware. This breakthrough not only showcases the technical prowess of Chinese engineers but also serves as a powerful reminder of the potential for innovation in resource-constrained environments.
DeepSeek's success is not an isolated phenomenon,but rather a reflection of the broader vibrancy of China's AI ecosystem. Other domestic AI models are also making significant strides in their respective domains. For example, Doubao by ByteDance is widely used in customer service and education; Kimi, a multimodal model, seamlessly integrates texts, images and audio files; Wenxin Yiyan by Baidu rivals ChatGPT in natural language understanding and generation; Tongyi Qianwen by Alibaba delivers tailored solutions for the e-commerce and logistics sectors with strong potential for localization.
These models collectively highlight the strength of China's AI industry and its ability to cater to the specific needs of domestic industries and consumers. Moreover, they underscore the importance of fostering a diverse and competitive AI ecosystem that can drive innovation on a global scale.
China's AI advancements are not confined to its borders, as they are increasingly making an impact on the global stage. Chinese AI technologies are expanding into markets in the Global South, where they are being integrated into healthcare, agriculture and education. By addressing local challenges and providing scalable solutions, these technologies are helping to bridge the digital divide and promote sustainable development.
The growing influence of Chinese AI also highlights the interconnectedness of global technological development. As China continues to innovate and overcome hardware constraints, it is expected to challenge the dominance of US artificial intelligence models and companies soon, fostering a more diverse, open, inclusive and fair competitive global AI ecosystem. This healthy competition is not only driving innovation but also accelerating technological progress, ultimately benefiting industries and consumers worldwide. That will ensure AI technologies serve as a shared resource, contributing to the construction of a more equitable and sustainable future.
China's ability to overcome hardware constraints and achieve significant advancements in AI underscores the limitations of technology embargoes as a tool to stifle innovation in the 21st century. While chip restrictions may temporarily slow progress, they have also spurred greater creativity. The story of DeepSeek and China's broader AI achievements is not just about technology; it is a testament to the enduring power of human ingenuity and the futility of attempts to contain it.
As the global AI landscape continues to evolve, China's contributions serve as a reminder of the importance of resilience, innovation and collaboration. By fostering an environment that encourages fair competition and shared progress, the international community can ensure that AI technologies are developed and deployed in ways that benefit all of humanity.
In this globalized context, the story of China's AI journey is not just a national narrative but a chapter in the broader story of human progress.
近年来，人工智能已经成为一股变革力量，重塑了行业，重新定义了全球技术格局。人工智能被广泛认为是第四次工业革命的基石，正在推动医疗保健、金融、制造业和运输等领域的突破。
这场革命的最前沿是一些总部位于美国的科技巨头，如英伟达、谷歌和英特尔，他们对研发的持续投资迄今为止巩固了他们在该领域的领导地位。围绕人工智能潜力的乐观情绪不仅推动了技术进步，还推动了相关公司股价的大幅上涨，激励了全球投资者。
话虽如此，全球人工智能竞赛并不是一场单方面的游戏。虽然美国长期以来一直是主导者，但中国已迅速成为一个强大的竞争对手，在面临重大挑战时表现出非凡的韧性和创新能力。其中最主要的挑战是美国的技术制裁，这限制了中国获得先进半导体芯片的机会。众所周知，这些高性能芯片对于训练人工智能模型至关重要。许多人认为，这种限制可能会严重阻碍中国在这方面的野心。然而，中国的人才出乎意料，利用独立研究、创意工程和战略投资来维持其在人工智能发展方面的势头。
由于高性能芯片的限制，面对先进计算能力的稀缺，中国工程师采用了创新的方法来优化人工智能工作流程。
模型修剪、量化和知识提取等技术已被用于减少人工智能模型的大小和复杂性，同时保持高精度，使中国人工智能能够在不太先进的硬件上高效运行，有效地绕过了对顶级芯片的需求。
例如，DeepSeek的设计是为了在Nvidia的H800芯片上表现出色，而H800芯片的性能不如ChatGPT等大型语言模型使用的顶级H100芯片。这一成就凸显了中国研究人员的聪明才智和适应受限环境的能力。
中国人工智能战略的另一个关键方面是国内硬件替代品的发展。华为等公司在设计和生产人工智能专用芯片方面取得了重大进展，如Ascend系列。尽管这些芯片的性能可能尚未达到国际同行的水平，但它们正在稳步缩小差距，为中国的人工智能雄心提供了可行的基础。对云计算和分布式计算基础设施的大量投资也是至关重要的，这些基础设施允许人工智能模型使用性能较低的处理器网络进行扩展。
中国作为人工智能强国的崛起，体现在其不断增长的尖端人工智能模型和技术组合上。其中，Deep Seek因其突破人工智能能力界限的能力而受到广泛关注，尽管存在硬件限制。与豆包、Kimi、文心一言、通义千问等国内人工智能模型一起，它们代表了中国在逆境中开辟独特发展道路的决心。
DeepSeek和其他中国人工智能模型的成功凸显了优化硬件和软件以实现人工智能突破的重要性。关注算法效率和利用国内硬件解决方案表明，即使在重大限制下，创新也能蓬勃发展。
DeepSeek是中国在人工智能领域创新能力的一个典型例子。基于先进的机器学习算法和广泛的数据集，它在自然语言处理、图像识别和预测分析方面表现出色。自第三次迭代发布以来，DeepSeek因其卓越的功能而获得认可，将自己定位为OpenAI ChatGPT的有力竞争者。
DeepSeek的独特之处在于，尽管在不太先进的硬件上运行，它仍能提供高性能。这一突破不仅展示了中国工程师的技术实力，也有力地提醒人们在资源受限的环境中进行创新的潜力。
DeepSeek的成功并不是一个孤立的现象，而是中国人工智能生态系统更广泛活力的反映。其他国内人工智能模型也在各自领域取得了重大进展。例如，字节跳动的豆包在客户服务和教育方面得到了广泛的应用；Kimi是一个多模式模型，无缝集成了文本、图像和音频文件；百度文心一言在自然语言理解和生成方面与ChatGPT竞争；阿里巴巴旗下的通义千问为电子商务和物流行业提供量身定制的解决方案，具有很强的本地化潜力。
这些模型共同突显了中国人工智能产业的实力及其满足国内行业和消费者特定需求的能力。此外，他们强调了培育一个多样化和有竞争力的人工智能生态系统的重要性，该生态系统可以在全球范围内推动创新。
中国的人工智能进步并不局限于国界，因为它们对全球舞台的影响越来越大。中国的人工智能技术正在向全球南方市场扩张，在那里它们正被整合到医疗保健、农业和教育领域。通过应对当地挑战和提供可扩展的解决方案，这些技术正在帮助弥合数字鸿沟，促进可持续发展。
中国人工智能日益增长的影响力也突显了全球技术发展的相互联系。随着中国不断创新和克服硬件限制，预计很快将挑战美国人工智能模型和公司的主导地位，培育一个更加多样化、开放、包容和公平竞争的全球人工智能生态系统。这种良性竞争不仅推动了创新，也加速了技术进步，最终使全球行业和消费者受益。这将确保人工智能技术成为一种共享资源，为建设一个更加公平和可持续的未来做出贡献。
中国克服硬件限制并在人工智能方面取得重大进展的能力突显了技术禁运作为21世纪扼杀创新的工具的局限性。虽然芯片限制可能会暂时减缓进展，但它们也激发了更大的创造力。DeepSeek和中国更广泛的人工智能成就的故事不仅仅与技术有关；这证明了人类创造力的持久力量，以及试图遏制它的徒劳。
随着全球人工智能格局的不断发展，中国的贡献提醒人们韧性、创新和协作的重要性。
在这种全球化的背景下，中国人工智能之旅的故事不仅是一个国家叙事，也是人类进步更广泛故事中的一章。
---语法填空改编---
In recent years, artificial intelligence has emerged as a transformative force, reshaping industries and redefining the global technological landscape. Widely regarded as a cornerstone of the Fourth Industrial Revolution, AI is driving breakthroughs in 1.____(field) such as healthcare, finance, manufacturing and transportation.
Faced with a scarcity in advanced computing power because of restrictions on high-performance chips, Chinese engineers have adopted innovative approaches 2.__(optimize) AI workflows.
Techniques such as model pruning, quantization and knowledge distillation 3._______(employ) to reduce the size and complexity of AI models while maintaining high levels of accuracy, enabling Chinese AI to operate 4._______(efficient) on less-advanced hardware, effectively bypassing the need for top-tier chips.Another critical aspect of China's AI strategy has been 5._______ development of domestic hardware alternatives. Companies such as Huawei have made significant strides 6.______designing and producing AI-specific chips, such as the Ascend series. While these chips may not yet match the performance of 7.______(they) international counterparts, they are steadily narrowing the gap and providing a viable foundation for China's AI ambitions. Substantial investments in cloud computing and distributed computing infrastructure, 8._______allow AI models to scale using networks of lower-performance processors, are also essential.
China's rise as an AI powerhouse 9._______(exemplify) by its growing portfolio of cutting-edge AI models and technologies. Among these, Deep-Seek has garnered significant attention for its ability to push the boundaries of AI capabilities despite hardware limitations. Alongside other domestic AI models, such as Doubao, Kimi, Wenxin Yiyan and Tongyi Qianwen, they represent China's 10._______(determine) to forge a unique developmental path in the face of adversity.
【参考答案】
1.fields 2.to optimize 3.have been employed 4.efficiently 5.the 6.in7.their 8.which 9.is exemplified 10.determination
---Part 2---
R1 exemplifies open spirit of the internet
[Photo/VCG]
The R1 model released by Chinese company DeepSeek has impressed those in the industry with all the features it shows.
Similar to OpenAI o1, the R1 model released by DeepSeek on Jan 20 achieves performance comparable at 3 percent of the cost. Deep-Seek R1 is a first-generation reasoning model, trained via large-scale reinforcement learning, like OpenAI o1.
However, unlike OpenAI o1, Deep-Seek R1 is totally open-sourced, allowing developers and professionals and everybody easy access to the codes it has used in developing the model.
Just like DeepSeek said on github.com, "The open source Deep-Seek-R1, as well as its API, will benefit the research community to distill better smaller models in the future".
Compared with the existing giants, DeepSeek, founded in 2023, is a young company. The small model that DeepSeek has chosen to develop is also a practical strategy for both itself and young enterprises like it in the industry. Currently the mainstream AI models describe things with parameters. For example they describe a banana with parameters that include "yellow","strip-shaped","edible" and "sweet".The number of parameters is an essential measurement of AI models, as the model that can accurately describe an object with fewer parameters has higher efficiency.
DeepSeek V3, one of the opensource models released by Deep-Seek, which shows higher efficiency,had only 671 billion parameters, of which 37 billion were activated in usage, both of which are quite low in the industry. More important, through deep exploration, it managed to train that model using only 2048 H800 GPUs of Nvidia at the cost of $5.6 million, which is just a small percentage of the cost incurred by OpenAI and Google for training similar types of models.
As the US, guided by its protectionist strategy,is still blocking exports of high-performance chips from its own and its allies' producers to China, the low cost and easy-to-obtain chips are essential for other Chinese developers. For young, innovative Chinese companies to grow into capable players on the global stage, that's the correct direction.
中国公司DeepSeek发布的R1型号以其展示的所有功能给业内人士留下了深刻印象。
与OpenAI o1类似，DeepSeek于1月20日发布的R1模型的性能仅为成本的3%。Deep Seek R1是第一代推理模型，通过大规模强化学习进行训练，如OpenAI o1。
然而，与OpenAI o1不同，Deep Seek R1是完全开源的，允许开发人员、专业人员和每个人轻松访问它在开发模型时使用的代码。
正如DeepSeek在github.com上所说，“开源的Deep-Seek-R1及其API将有利于研究社区在未来提取更好的小型模型”。
与现有的巨头相比，成立于2023年的DeepSeek是一家年轻的公司。DeepSeek选择开发的小型模式对其自身和行业中的年轻企业来说都是一种实用的策略。目前主流的人工智能模型用参数来描述事物。例如，他们用包括“黄色”、“条形”、“可食用”和“甜味”在内的参数来描述香蕉。参数的数量是人工智能模型的重要衡量标准，因为能够准确描述参数较少的对象的模型具有更高的效率。
DeepSeek V3是Deep Seek发布的开源模型之一，显示出更高的效率，只有6710亿个参数，其中370亿个参数在使用中被激活，这两个参数在行业中都相当低。更重要的是，通过深入探索，它仅使用英伟达的2048个H800 GPU就以560万美元的成本训练了该模型，这只是OpenAI和谷歌训练类似类型模型所产生成本的一小部分。
由于美国在其保护主义战略的指导下，仍在阻止其本国及其盟友的生产商向中国出口高性能芯片，因此低成本和易于获得的芯片对其他中国开发商至关重要。对于年轻、有创新精神的中国公司来说，在全球舞台上成长为有能力的参与者，这是正确的方向。
---语法填空改编---
The R1 model released by Chinese company DeepSeek has impressed those in the industry with all the features it shows.Similar to OpenAI o1, the R1 model 1.______(release) by DeepSeek on Jan 20 achieves performance comparable at 3 percent of the cost. Deep-Seek R1 is 2.______first-generation reasoning model, trained via large-scale reinforcement learning, like OpenAI o1.
However, unlike OpenAI o1, Deep-Seek R1 is totally open-sourced, 3._____(allow) developers and professionals and everybody easy access to the codes it _____(use) in developing the model.Just like DeepSeek said on github.com, "The open source Deep-Seek-R1, as well as its API, will benefit the research community to distill better 5.______(small) models in the future".
Compared with the existing giants, DeepSeek, founded in 2023, is a young company. The small model that DeepSeek has chosen to develop is also a practical strategy for both 6.______(it) and young enterprises like it in the industry. Currently the mainstream AI models describe things with parameters. DeepSeek V3, one of the opensource 7.______(model) released by Deep-Seek, which shows higher efficiency, had only 671 billion parameters, of which 37 billion 8._____(activate) in usage, both of which are quite low in the industry. More important, through deep exploration, it managed to train that model using only 2048 H800 GPUs of Nvidia 9._______the cost of $5.6 million, 10._______is just a small percentage of the cost incurred by OpenAI and Google for training similar types of models.
【参考答案】
1.released 2.a 3.allowing 4.has used 5.smaller 6.its 7.models 8.were activated 9.at 10.which
声明：转载请完整备注“晓予说”。本时文英语原文和图片来自China Daily，语法填空由小予君改编。如有侵权请联系删除。
