DeepSeek AI: 5 Myths and Facts About China’s Fastest-Growing AI Startup

Think you know everything about DeepSeek, China’s rising AI powerhouse? Think again! 🤯 Here are five common misconceptions about this AI startup—debunked with the real facts. From its true innovation power to its global impact, let’s separate myth from reality. 🔍 #AI #DeepSeek #TechTrends

DeepSeek has received much praise from both tech industry executives and lawmakers following the release of the Chinese AI startup’s reasoning model R1, which prompted a wider sell-off in tech stocks across markets from New York to Tokyo. The quality and cost-efficiency of its R1 model is what has primarily led to DeepSeek’s surge in popularity.

The company has claimed that its AI model matches with, and in some cases beats, OpenAI’s o1 reasoning model in performance while using fewer graphics processing units (GPUs) and costing far less. The AI chatbot app DeepSeek, which gives free access to R1, has topped app store charts in multiple countries. But the success story of the company has also drawn caution and scepticism from a few. OpenAI has accused DeepSeek of IP theft and said it has evidence of the company using its GPT models to train its own.

Myth #1-DeepSeek’s AI models signal AGI is within reach

Reality- : DeepSeek’s AI models are a big improvement in efficiency and cost but do not necessarily signal a step toward artificial general intelligence (AGI).

AGI is a term used by the tech industry to describe an AI model capable of equaling or surpassing human intellect on a wide range of tasks. No one has declared that they have developed such an AI model yet. However, OpenAI and some of its rivals have said that they are eagerly working towards reaching the AGI milestone.

In 2023, DeepSeek transformed from a research unit at Chinese hedge fund High-Flyer’s AI research unit to becoming an AI company. The firm was founded by hedge fund manager Liang Wenfeng, whose goal is the development of LLMs as part of a path to achieving AGI

Although R1 marks a turning point in the race of AI supremacy, DeepSeek came up with no new technology. “Getting to AGI probably requires five or six more breakthroughs and the company or country that can ramp up those breakthroughs first may win,” said Professor Gary Marcus at New York University (NYU), who is an AI expert while talking to CNBC.

Myth #2: DeepSeek’s breakthrough shows export controls don’t work

Reality: The US export controls on the sale of high-end GPUs could also affect China’s AI development for a long period.

DeepSeek was a breakthrough due to the ‘unintentional’ ramification of a US export-control measure that thwarted Chinese high-tech firms’ efforts to scale their AI models by buying higher-end GPUs available in the country. Since their access to those top-of-line chips from the US-based NVIDIA company was lost, researchers engaged in DeepSeek were reportedly asked to come up with clever ‘workarounds’ to boost the efficiency levels of AI model consumption of the raw compute powers.

Critics have argued that US export controls backfired, but DeepSeek reportedly stockpiled 10,000 of Nvidia’s older generation A100 GPUs before the trade restrictions were imposed. Miles Brundage, an AI policy expert who recently left OpenAI, has suggested that export controls might still slow China down when it comes to running more AI experiments and building AI agents.

As explained by Brundage during the latest podcast: “DeepSeek was pushed from necessity to search for some of those techniques, perhaps quicker than US-based corporations would have otherwise. However, that does not equate to mean that they mightn’t like to have that many [GPUs] that are certainly quite far away in order for them to start instantly transition from o1 to o3 or from o1 to o5 similarly the Open AI transition did have to its massive chip fleet “.

But perhaps even more consequential, Dario Amodei, CEO of Anthropic-the company that had developed the Claude series of AI models the DeepSeek’s results “makes export control policies even more existentially important than they were a week ago

Myth #3: DeepSeek is seriously a threat to Nvidia

Reality: DeepSeek’s R1 model is not as dangerous for Nvidia after all.

A flurry of hype over DeepSeek sent the investors of Nvidia into a panic, where the stock lost 17% and almost wiped out $600 billion in market value on January 27. The stock of the chip giant recovered from the sharp fall on January 28, it fell another 4 per cent on January 29. While DeepSeek’s R1 model may have reduced the need for massive collections of special-purpose AI hardware from the likes of Nvidia, it does not exactly mean the end of the road for the chip giant.

Microsoft CEO Satya Nadella noted that DeepSeek’s influence could, paradoxically, boost demand for high-end GPUs. “Jevons paradox strikes again!” Nadella wrote in a post on X. Jevons Paradox is an economic theory which suggests that when technological progress makes the use of a resource more efficient, overall consumption of that resource tends to increase. Tech investor Andrew Ng said that how DeepSeek’s results will diminish demand for the necessary GPUs and compute power has yet to be seen. “Sometimes making each unit of a good cheaper can result in more dollars in total going to buy that good,” he said in a post on X.

Myth #4: DeepSeek R1 is an all-open model

Reality: DeepSeek R1 is free to download, modify, and reuse, but it may not be considered truly open-source.

The impressive results of DeepSeek R1 have been interpreted by many as a sign that China is pulling ahead of the US in the race for AI supremacy. But beyond the geopolitical angle, DeepSeek’s success is also being celebrated as a win of open-source AI over closed AI.

, according to this, Meta’s chief AI scientist, Yann LeCun said, “DeepSeek has benfitted from open research and open source (e.g., PyTorch and Llama from Meta). They developed new ideas and constructed them on top of other people’s work. Because their work is published and open source, everyone can benefit from it. That is the power of open research and open source.”

The model architecture and the weights that have been used by R1 were made public under a permissive MIT licence. This allows the deployment of the model with no restrictions. However, R1 does not satisfy the generally accepted definition of what ‘open-source’ means. Indeed, OSI claims that it should be open source if and only if an open-source AI model ensures making the details of the data used for training the AI, the full code used for its conception and running, as well as the settings and weights from the training available.

The training data of R1 have never been disclosed. The program and other instructions for training do not exist also. There are strong resistance from the AI open-source developers to disclose any training data in the first place because there exists a legitimate prospect of winning litigation on copyright violation.

Myth #5: DeepSeek AI models pose a bigger privacy risk

Reality: AI models of DeepSeek pose identical privacy risk than other LLMs

DeepSeek’s meteoric rise in popularity has coincided with concerns related to the data privacy of users and authorities. Some such concerns came with the Chinese origins of the AI startup, while others complained that its AI technology was open-source.DeepSeek clearly states, under the privacy policy: “We store the information we collect in secure servers located in the People’s Republic of China.”.

Nevertheless, tech industry figures like Perplexity CEO Aravind Srinivas have repeatedly looked to assuage such data protection concerns by pointing out that R1 model can be downloaded and run locally on your laptop or other devices. Running local instances means users can privately interact with DeepSeek‘s AI without the company getting their hands on input data.

Conclusion

DeepSeek AI has captured global attention with its powerful R1 model, but its rise is also surrounded by myths and misconceptions. While its innovations in cost-efficiency and reasoning capabilities are impressive, claims about AGI breakthroughs, export control impacts, and privacy risks require deeper scrutiny. As the AI landscape continues evolving, DeepSeek’s influence will be a key factor to watch—but separating fact from fiction remains crucial. Stay informed, stay curious, and keep exploring the world of AI!