DeepSeek issues an open challenge

Unless you were hiding under a rock, you must be familiar with the name DeepSeek by now. This Chinese company has set the tech world alight by unveiling an artificial intelligence model that challenges some of the best in the market but built at a much lower cost. Moreover, they released it as an open-source model, allowing users worldwide free access. Previously, AI models from big tech players like OpenAI and Google required users to shell out thousands of dollars.

This predictably set off a panic among fickle minded investors who were backing big spenders like OpenAI, Google and Meta and high end chip maker Nvidia, counting on their unrestrained growth and market dominance. While the jury is still out on the impact of DeepSeek on the sector per se, there is no dispute that US-based AI companies are going to face some searching questions about their high cost. 

The Chinese firm says it took only two months for them to build R1 and cost them less than 6 million US dollars. Some dispute this claim, but even if they have spent ten times more, it would be still cheaper than what others spent.

The cost to users comes down dramatically if they opt for DeepSeek. The Chinese company charges just over 2 US dollars per 1 million output tokens, compared to 60 US dollars for GPT-o1. Even OpenAI's mini version of GPT-o1, which is supposed to be more affordable version, costs 12 US dollars. This could lower the entry barrier to thousands of companies.

This has already made an impact and DeepSeek has become the most downloaded app in 140 countries and India heads the list, accounting for 15.6 per cent of all downloads across platforms since the app’s launch in January.

In China, this was nothing new as DeepSeek has done this before. Back in November 2023, Deep Seek introduced their V2 models which were priced lower, forcing other Chinese companies like Bytedance and Alibaba also to lower their rates.

But things truly escalated in December 2024 when DeepSeek launched their V3 models and rolled out their R1 reasoning models that performed on par with OpenAI’s 01 model. Then everything hit the fan.

Deep Dive

Their models were scrutinised intensely and many found faults, most of which you can slap on any Chinese model, like censorship. An analysis by the MIT Technology Review looked at their work and found the Chinese company optimised what was available, like making reinforcement learning efficient through automation rather than trying OpenAI’s method of using human feedback to make it slick.  

“On the hardware side, DeepSeek found new ways to juice old chips, allowing it to train top-tier models without coughing up for the latest hardware on the market. Half their innovation comes from straight engineering,” says founder of AI firm Claifai, Matt Zeiler, says in the report that goes in-depth into DeepSeek methods.

The shock waves that DeepSeek sent through Silicon Valley was natural as the narrative till now has been that AI is a hugely expensive affair. This along with restrictions on sale of high end Nvidia chips meant Chinese models would struggle to catch up with the leaders like OpenAI, said US-based analysts. 

What China has been showing consistently was that they can take a technology and tweak it into better products. That had made the country the second biggest economy in the world.

It may not be an apple to apple comparison here, but developments in other sectors support this. Take the case of Temu and Shein in e-commerce, BYD in electric car manufacture, social media app WeChat and heavy industries like ship building and bullet trains – China has been innovating their way to the top steadily. Even when restrictions were put on Huawei, that did not stop the phone company from coming out with a cutting edge Mate 60 phones and develop their own operating system, Harmony.

More Than Money

But DeepSeek founder Liang Wenfeng’s vision extends beyond mere commercial success. In an interview given in November 2024, he articulated a philosophy that perhaps explains DeepSeek’s approach. “China should gradually become a contributor instead of freeriding,” he said, reflecting on how China had largely been a consumer of Western technological innovation for the past three decades.

“We’re used to Moore’s Law falling out of the sky, lying at home waiting 18 months for better hardware and software to emerge. That’s how the Scaling Law is being treated. But in fact, this is something that has been created through the tireless efforts of generations of Western-led tech communities.”

“What we see is that Chinese AI can't be in the position of following forever. We often say that there is a gap of one or two years between Chinese AI and the United States, but the real gap is the difference between originality and imitation,” he says in another interview.

DeepSeek is what OpenAI was a few years back when the American firm was pursuing pure research before it was reshaped into a commercial entity. Liang and his team made up of mostly young engineers have no pressure to show profitability and continue to work on their pursuit of artificial general intelligence (AGI).

What they will have trouble ahead is when the Chinee Communist Party leadership start squeezing them. “Now that folks on the Central Committee are going to be following this more closely than before, this open versus closed source dynamic with AI software – I see downside scenarios for a company that really prides itself on putting everything it does out into the public,” wrote Jordan Schneider, founder of China Talk.

Indian Research

The contrasting trajectories of AI development in India and China, exemplified by DeepSeek’s success, highlight fundamental differences in approach and ecosystem. While Silicon Valley leaders like Sam Altman discouraged competition in foundation models, suggesting it was “hopeless” to compete with established players, DeepSeek’s achievements in China have challenged this narrative by demonstrating that smaller teams with modest resources can indeed develop competitive AI models.

In India, most AI companies and influencers stuck to it was too expensive narrative. It was even backed by tech influencerslike Nandan Nilekani, who says practical applications and infrastructure development are the priorities over fundamental research.

This stands in sharp contrast to China’s ecosystem, where DeepSeek represents just one of many ambitious AI research projects. The patent statistics further illuminate this divide.While India has seen significant growth in patent grants (from 8,248 in 2016 to 76,053 in 2023), startup participation in patent applications remains surprisingly low at just 6.3 percent, suggesting a persistent focus on immediate commercial applications rather than foundational research.

A new world is shaping up

Out goesthe boss, in comes the family. That is the priority of Indian workforce, according to a survey done by Indeed's Future Career Resolution. The trend suggests a possible post-pandemic recalibration of work-life balance, though interestingly, this shift coexists with strong optimism about job prospects, particularly in emerging tech sectors, where 55 percent see growth potential. The survey also says Workers are investing in “future-proof” skills such as data literacy, machine learning and emotional intelligence to stay competitive. Meanwhile, nearly six in 10 believe hiring will focus more on demonstrable skill sets than formal degrees, a trend that’s aligned with rapid shifts in tech-driven sectors.

Over in the US, a survey finds teens are losing trust in Big Tech, marking a generational turning point. Having grown up as digital natives, these teens’ scepticism, particularly regarding AI ethics and data privacy, suggests a more sophisticated and critical understanding of technology's impact. The fact that nearly half express distrust in tech companies’ AI decisions is especially telling, as it indicates that younger users are increasingly aware of and concerned about the broader implications of technological advancement.

AI to tackle missing children cases

Thousands of children go missing in India every day and finally technology is put to use to prevent this. Khoji.in is an innovative application of AI technology to address this serious social issue in India as an estimated 96,000 children go missing annually. The platform’s integration with DigiLocker and its government-patented one-click identification system demonstrates how targeted AI solutions can address specific societal challenges. The father-son founding team’s approach to this problem shows how tech entrepreneurship can be directed toward social impact rather than just commercial success, says YourStory.

Robots put on the dancing shoes

Robots as a coworker has been warned about. But as dance partners? In what might be the world’s most unexpected dance crew, 16 Unitree robots decked out in snazzy red-and-white vests strutted their stuff at the Spring Festival Gala last week, China's biggest Television spectacle. These mechanical movers shimmied alongside human dancers in a Yangge folk dance routine, powered by AI that supposedly helped them “understand” the music. With over a billion viewers watching, these metallic maestros proved that “dancing like a robot” could actually be a compliment.