DeepSeek Uncovered: 5 Myths and Truths Behind the Chinese AI Startup’s Rise
The hype surrounding DeepSeek is starting to face reality, as questions emerge about how its AI models were built and what they mean for the industry.
Just days after launching its reasoning model, R1, DeepSeek has been praised by tech leaders and policymakers alike. But its success has also shaken up the market, triggering a tech stock sell-off from New York to Tokyo.
.jpeg "DeepSeek Uncovered: 5 Myths and Truths Behind the Chinese AI Startup’s Rise")
The reason? R1 delivers high-quality AI reasoning at a fraction of the cost, outperforming OpenAI’s o1 model in some cases—while using fewer GPUs. DeepSeek’s chatbot app, which provides free access to R1, has skyrocketed to the top of app store charts in multiple countries.
But with rapid success comes scrutiny. OpenAI has accused DeepSeek of stealing intellectual property, claiming it has proof that DeepSeek trained its models using OpenAI’s GPT technology.
As debates heat up, it’s time to separate fact from fiction. Here are five common myths about DeepSeek’s rise—and the truth behind them.
Myth #1: DeepSeek’s AI proves AGI is within reach
Reality: While R1 is a major leap in efficiency, it doesn’t mean we’re close to Artificial General Intelligence (AGI).
AGI refers to AI that can match or surpass human intelligence across a wide range of tasks. Despite the hype, no company—including DeepSeek—has cracked AGI yet. OpenAI and others are chasing this goal, but it remains out of reach.
DeepSeek, which evolved from a hedge fund’s AI research unit in 2023, was founded by Liang Wenfeng with the ambition of developing large language models (LLMs) that could eventually lead to AGI.
OpenAI CEO Sam Altman acknowledged R1’s impact, calling it “impressive,” but he quickly shifted the conversation back to AGI—suggesting that the real breakthrough is still ahead.
Gary Marcus, an AI expert and NYU professor, believes we’re still five or six major breakthroughs away from AGI. The race is far from over.
Reality: While R1 is a major leap in efficiency, it doesn’t mean we’re close to Artificial General Intelligence (AGI).
AGI refers to AI that can match or surpass human intelligence across a wide range of tasks. Despite the hype, no company—including DeepSeek—has cracked AGI yet. OpenAI and others are chasing this goal, but it remains out of reach.
DeepSeek, which evolved from a hedge fund’s AI research unit in 2023, was founded by Liang Wenfeng with the ambition of developing large language models (LLMs) that could eventually lead to AGI.
OpenAI CEO Sam Altman acknowledged R1’s impact, calling it “impressive,” but he quickly shifted the conversation back to AGI—suggesting that the real breakthrough is still ahead.
Gary Marcus, an AI expert and NYU professor, believes we’re still five or six major breakthroughs away from AGI. The race is far from over.
Myth #2: DeepSeek’s success proves US export controls have failed
Reality: While DeepSeek found ways to innovate despite restrictions, US export controls still pose a challenge for China’s AI growth.
DeepSeek’s rise is often cited as proof that US efforts to block China’s access to advanced GPUs backfired. However, the reality is more nuanced.
Before US restrictions took effect, DeepSeek reportedly stockpiled 10,000 Nvidia A100 GPUs—older but still powerful chips. While the company has optimized its AI models to be more efficient, it remains unclear whether China can continue advancing AI at the same pace without access to cutting-edge hardware.
Miles Brundage, an AI policy expert, argues that while DeepSeek found creative workarounds, it still lacks the resources to scale as rapidly as OpenAI, which has access to a much larger fleet of advanced chips.
Anthropic CEO Dario Amodei reinforced this point, saying DeepSeek’s success makes strict export controls “more critical than ever.”
Reality: While DeepSeek found ways to innovate despite restrictions, US export controls still pose a challenge for China’s AI growth.
DeepSeek’s rise is often cited as proof that US efforts to block China’s access to advanced GPUs backfired. However, the reality is more nuanced.
Before US restrictions took effect, DeepSeek reportedly stockpiled 10,000 Nvidia A100 GPUs—older but still powerful chips. While the company has optimized its AI models to be more efficient, it remains unclear whether China can continue advancing AI at the same pace without access to cutting-edge hardware.
Miles Brundage, an AI policy expert, argues that while DeepSeek found creative workarounds, it still lacks the resources to scale as rapidly as OpenAI, which has access to a much larger fleet of advanced chips.
Anthropic CEO Dario Amodei reinforced this point, saying DeepSeek’s success makes strict export controls “more critical than ever.”
Myth #3: DeepSeek’s R1 model is a major threat to Nvidia
Reality: While R1 is a game-changer, it’s not a death sentence for Nvidia.
After R1’s launch, Nvidia’s stock took a hit—dropping 17% and wiping out nearly $600 billion in market value on January 27. The stock partially recovered, but concerns lingered.
The fear? If AI models become more efficient, companies may need fewer GPUs, cutting into Nvidia’s dominance.
However, Microsoft CEO Satya Nadella sees it differently. He pointed out that efficiency often increases overall demand—a phenomenon known as Jevons Paradox. If AI becomes more accessible, more people will want to use it, potentially increasing demand for GPUs rather than reducing it.
Tech investor Andrew Ng echoed this view, arguing that making AI cheaper could actually lead to higher overall spending on compute power. The jury’s still out on how this will play out for Nvidia.
Reality: While R1 is a game-changer, it’s not a death sentence for Nvidia.
After R1’s launch, Nvidia’s stock took a hit—dropping 17% and wiping out nearly $600 billion in market value on January 27. The stock partially recovered, but concerns lingered.
The fear? If AI models become more efficient, companies may need fewer GPUs, cutting into Nvidia’s dominance.
However, Microsoft CEO Satya Nadella sees it differently. He pointed out that efficiency often increases overall demand—a phenomenon known as Jevons Paradox. If AI becomes more accessible, more people will want to use it, potentially increasing demand for GPUs rather than reducing it.
Tech investor Andrew Ng echoed this view, arguing that making AI cheaper could actually lead to higher overall spending on compute power. The jury’s still out on how this will play out for Nvidia.
Myth #4: DeepSeek R1 is fully open-source
Reality: While R1 is freely available, it doesn’t meet the strict definition of open-source AI.
R1’s open access has been hailed as a win for open-source AI, especially as it challenges the closed models of companies like OpenAI.
Meta’s chief AI scientist, Yann LeCun, praised DeepSeek for building on open research, noting that the company used innovations from projects like PyTorch and Llama. But while R1 can be downloaded and modified for free, it doesn’t meet the Open Source Initiative’s (OSI) strict definition of open-source.
To be truly open-source, AI models must share their training data, code, and full training processes. DeepSeek has not released its training data or detailed how R1 was built—likely to avoid potential copyright and legal issues.
So while R1 is more open than most AI models, it’s not fully open-source in the traditional sense.
Reality: While R1 is freely available, it doesn’t meet the strict definition of open-source AI.
R1’s open access has been hailed as a win for open-source AI, especially as it challenges the closed models of companies like OpenAI.
Meta’s chief AI scientist, Yann LeCun, praised DeepSeek for building on open research, noting that the company used innovations from projects like PyTorch and Llama. But while R1 can be downloaded and modified for free, it doesn’t meet the Open Source Initiative’s (OSI) strict definition of open-source.
To be truly open-source, AI models must share their training data, code, and full training processes. DeepSeek has not released its training data or detailed how R1 was built—likely to avoid potential copyright and legal issues.
So while R1 is more open than most AI models, it’s not fully open-source in the traditional sense.
Myth #5: DeepSeek’s AI models pose a greater privacy risk
Reality: DeepSeek’s privacy risks are similar to those of other large language models.
Concerns about DeepSeek’s AI often stem from its Chinese origins. Some fear that because the company is based in China, user data could be at risk.
DeepSeek’s privacy policy states that its servers are located in China, which has fueled these concerns. However, AI industry leaders argue that DeepSeek’s R1 model can be run locally, meaning users don’t have to send their data to DeepSeek’s servers at all.
Perplexity CEO Aravind Srinivas pointed out that his company is hosting R1 in the US and Europe, ensuring that users can interact with the model outside of China’s jurisdiction.
While privacy concerns are valid for any AI model, DeepSeek’s risks appear to be on par with those of other major AI players.
Reality: DeepSeek’s privacy risks are similar to those of other large language models.
Concerns about DeepSeek’s AI often stem from its Chinese origins. Some fear that because the company is based in China, user data could be at risk.
DeepSeek’s privacy policy states that its servers are located in China, which has fueled these concerns. However, AI industry leaders argue that DeepSeek’s R1 model can be run locally, meaning users don’t have to send their data to DeepSeek’s servers at all.
Perplexity CEO Aravind Srinivas pointed out that his company is hosting R1 in the US and Europe, ensuring that users can interact with the model outside of China’s jurisdiction.
While privacy concerns are valid for any AI model, DeepSeek’s risks appear to be on par with those of other major AI players.
Final Thoughts:
DeepSeek’s rise is a wake-up call for the AI industry. It proves that AI innovation isn’t limited to Silicon Valley and that China is capable of competing at the highest level.
But while R1 is a breakthrough, it hasn’t rewritten the rules of AI—at least not yet. The race toward AGI is still ongoing, and the long-term impact of DeepSeek’s success remains to be seen.
As AI continues to evolve, one thing is clear: The competition is only getting fiercer.
.jpeg&description=DeepSeek Uncovered: 5 Myths and Truths Behind the Chinese AI Startup’s Rise){kind=link}