Loading Now

With China’s DeepSeek and Qwen, has AI war reached a tipping point? – Firstpost

With China’s DeepSeek and Qwen, has AI war reached a tipping point? – Firstpost



The artificial intelligence (AI) war has reached a tipping point, with DeepSeek firing a salvo at America’s Wall Street —either bursting the hype or replacing it with a new one. With China’s DeepSeek and now even Alibaba’s Qwen sending disruptive shockwaves across the AI landscape, it is essential to critically dissect their capabilities, limitations, and broader impact. We must assess them through the lenses of technological merit, long-term feasibility, geopolitics, data, and socio-political ramifications.

However, there is one disclaimer: current AI models, including OpenAI’s, are more on hype than actual performance. The hype of hope is that they will be able to do more than process words and produce actual insights instead of just arranging one word after the other based on their vast word repository.

Alibaba has introduced the Qwen series of AI models, with the latest being Qwen 2.5-Max. Alibaba claims Qwen 2.5-Max outperforms leading models from DeepSeek, OpenAI, and Meta platforms in key benchmarks. This model is designed to handle various tasks, including natural language understanding, mathematical problem-solving, and coding.

OpenAI claims that DeepSeek has used a distillation technique on its model to build its models. DeepSeek focuses on cost-effective development, minimising computational resources, and promoting open-source accessibility per the Chinese definition of open source. In contrast, Alibaba employs a hybrid model of proprietary and open-source.

One critical question surrounding DeepSeek is its access to data. While OpenAI and Meta’s LLAMA face legal scrutiny in securing large, high-quality datasets due to privacy and ethical constraints, DeepSeek benefits from China’s expansive and less restrictive data policies. China has access to vast pools of internet content, government-controlled data sources, and large-scale user interactions, giving it an edge in AI model training. If DeepSeek has leveraged unique proprietary datasets, this could explain its rapid progress compared to its Western counterparts.

DeepSeek has likely scraped all of the internet data, much like its Western counterparts, to train its models. While technically feasible, this practice raises legal and ethical concerns. Given the increasing regulatory scrutiny in the US and Europe over AI companies scraping copyrighted or proprietary data, DeepSeek could face similar legal challenges if deployed globally.

Whether China’s regulatory environment shields it from such legal hurdles remains to be seen, but international concerns about data privacy and intellectual property could hinder its global acceptance. Moreover, a new regulatory shield is likely to emerge faster in the Western world, similar to how China splintered the internet through censorship and restricted access.

DeepSeek may or may not represent a true scientific breakthrough but rather an incremental improvement accompanied by extensive global narrative building. This narrative has already destroyed and wiped out $1.2 trillion in wealth from Wall Street.

Unprecedented power

The geopolitical implications of DeepSeek or Qwen extend far beyond technological innovation. AI supremacy has become a critical component of global power, and China’s investment in AI is a strategic manoeuvre to challenge the dominance of the US and Western tech firms. With AI increasingly integrated into defence, cybersecurity, and economic planning, DeepSeek’s development signals China’s ambitions to lead the next technological revolution.

Concerns over data privacy, intellectual property, and the potential use of AI for surveillance or military applications will likely shape the international reception of DeepSeek. Countries wary of China’s influence may impose regulatory barriers, restricting DeepSeek’s adoption outside its home market. Additionally, DeepSeek’s success could push Western nations to accelerate their own AI advancements, intensifying the AI arms race.

China’s AI efforts, while formidable, will now face more obstacles, such as regulatory scrutiny, limited access to cutting-edge semiconductor hardware, and challenges in aligning models with human values. If DeepSeek is developed under restrictions that limit access to critical computational resources, its scalability may be hampered. More importantly, scrutiny over the transparency of DeepSeek’s AI methodology and its alignment with global ethical standards will now begin in earnest.

Model size, data curation, fine-tuning strategies, and real-world generalisation capabilities all limit AI advancements. AI models trained on vast amounts of data often exhibit shallow generalisation, struggling with out-of-distribution scenarios. If DeepSeek follows the same trajectory, its robustness remains questionable. Furthermore, while performance on benchmarks may seem impressive, real-world deployment introduces challenges in bias mitigation, ethical considerations, and reliability.

What racial, data, and cultural biases will DeepSeek generate and propagate across the internet? How will the Western world respond with regulations, controls, and laws to contain it? This is the next salvo in the AI war.

The global AI race is as much about geopolitics as it is about technology. DeepSeek represents a strategic move by China to establish AI dominance—arguably the most significant salvo fired by China in the AI arms race, where geopolitical motives often outweigh ethical considerations.

If DeepSeek indeed introduces a paradigm shift, one must ask: is it aligned with the principles of beneficial AI? Or does it introduce novel risks, such as enhanced deception, deepfake realism, or autonomous weaponisation?

Furthermore, interpretability becomes crucial as AI progresses towards artificial general intelligence (AGI). Without clear insight into DeepSeek’s decision-making processes, it remains another “black box” system—a concern often raised by Hinton.

DeepSeek, like many AI projects before it, walks a fine line between genuine innovation and overhyped potential. While its existence signifies China’s determination to lead in AI, its fundamental breakthroughs remain to be validated. If DeepSeek’s contributions are more about strategic positioning than scientific novelty, its impact may be more political than technological.

AI’s next frontier is not just about larger models but smarter, more interpretable, and ethically aligned ones. DeepSeek has further splintered the AI landscape, dividing it into a Chinese one versus the rest of the Western world.

There are important lessons for 
India here. First, individual, corporate, and even national data privacy must be safeguarded against AI-driven exploitation. Second, building AI models is not merely about throwing capital into innovating the parameters; structuring the model effectively is equally important. Third, with rising global scrutiny, the Indian government must take stronger measures to prevent the emergence of monopolistic AI dominance.

There will be more announcements from Chinese AI companies about better, faster, and more efficient AI engines, as the announcement itself has now become a weapon of mass destruction.

K Yatish Rajawat is a public policy researcher and works at the Gurgaon-based think and do tank Centre for Innovation in Public Policy (CIPP). Views expressed in the above piece are personal and solely that of the author. They do not necessarily reflect Firtspost’s views.



Source link

Post Comment