Deepseek: A Revolution Or Just A Chinese Propaganda Machine?

Two years ago, when tech giants like Microsoft and OpenAI decided to drop the biggest bombshell of the decade, generative AI, the world collectively lost its mind.

One day, people were content Googling answers to basic questions; the next, AI was writing their essays, coding their websites, generating their happy moments and even pretending to be their therapist. It was all very impressive, slightly terrifying and undeniably entertaining.

As expected, Google, perhaps realising it had been caught napping, hastily threw its hat into the ring, launching Bard (later rebranded as Gemini, because clearly, a name change fixes everything).

Meanwhile, everyone assumed that China was light-years behind in the AI race from the USA, busy making budget smartphones full of bloatware and surveillance cameras instead of groundbreaking AI models.

But here’s the thing, China wasn’t twiddling its thumbs and overthinking. Companies like Alibaba and Baidu were burning through cash faster than the divorce settlements of cricketers, desperately trying to develop an AI that could rival ChatGPT.

The failure? Despite all that frantic spending, their models landed with all the impact of a deflated balloon. People yawned, shrugged and went back to using GPT-4.

And just when it seemed like China’s AI ambitions were destined to be nothing more than expensive science projects, along came DeepSeek, the underdog no one saw coming. In a plot twist worthy of a Hollywood drama, this Chinese AI firm managed to do what its wealthier competitors couldn’t: build an AI model that actually impressed people.

Enter DeepSeek-R1 and DeepSeek-R1 Zero, the latest AI models that have left the internet collectively shaking their heads in the negative zone. These models aren’t just good—they’re terrifyingly good.

They offer state-of-the-art performance while costing a fraction of what OpenAI and Google charge. Suddenly, the AI landscape isn’t just a two-horse race anymore.

And the best part? The same tech enthusiasts who had written off China’s AI efforts as laughable failures are now tripping over themselves to praise DeepSeek. Oh, how the tables have turned, left to right or maybe upside down.

What is DeepSeek?

DeepSeek was founded in the last quarter of 2023 by Liang Wenfeng in Hangzhou. He is renowned as a serial entrepreneur with an impressive history in technology and investment. In 2013, he also co-founded Hangzhou Jacobi Investment Management, which employed AI to implement trading strategies, alongside fellow alumni of Zhejiang University.

DeepSeek R1 is an advanced language model built on a hybrid architecture and is significantly faster than its predecessor, v3. While DeepSeek is not the first AI model developed in China, it is considered more advanced than other AI models, including ChatGPT and Gemini. It incorporates large-scale reinforcement learning (RL) and chain-of-thought reasoning to enhance final outputs.

Sam Altman of OpenAI shares a similar vision of creating a model that surpasses human intelligence or achieves superhuman levels of perfection.

Why is DeepSeek in the spotlight?

The main reason for DeepSeek's increasing attention is its cost-effectiveness. To provide a quick comparison, OpenAI charges USD 15 per million input tokens and USD 60 per million output tokens. In contrast, DeepSeek offers a substantially lower price of USD 0.55 per million input tokens and USD 2.19 per million output tokens.

This makes it an attractive option for developers, researchers and organisations seeking cost-efficient AI solutions. In addition to being budget-friendly, DeepSeek has outperformed OpenAI in several tasks, including coding, where it has achieved a remarkable 97 per cent success rate.

Furthermore, DeepSeek has introduced six additional models, ensuring smooth operation on a variety of devices, including lower-powered laptops. While OpenAI, Microsoft and Google have invested billions of dollars into AI development, DeepSeek has reportedly only invested USD 6 million.

Is China getting better at AI, or is it just propaganda?

NVIDIA recently lost USD 20 billion in a single day, marking its biggest drop since March 2020. Reports suggest that DeepSeek’s cost efficiency is a major advantage but since these reports originate from China, there is speculation that they might be aimed at attracting investment. However, DeepSeek’s accuracy appears to speak for itself, often outperforming OpenAI and other AI models.

https://www.instagram.com/nikxopinion/reel/DFZnp2WRbay/

Despite the hype, the AI platform has already been caught up in controversy. When users search for terms such as "Arunachal Pradesh" or "North India", DeepSeek responds with, "Let’s talk about something else." This has sparked outrage, particularly among Indians, as Arunachal Pradesh is an integral part of India. China has repeatedly attempted to claim the territory but the Indian Army has successfully defended it.

A similar situation occurred with Chinese tech giants Huawei and Xiaomi, which were found to be collecting user data, leading to bans in several countries and states. There are concerns that DeepSeek could follow a similar trajectory. Meanwhile, US tech companies continue to advance rapidly.

As a result, the hype surrounding DeepSeek could ultimately prove to be a temporary bubble.