DeepSeek: Thief in the Night or a Harbinger of AI’s Future?

Is DeepSeek a threat to US AI?

The global artificial intelligence (AI) landscape is undergoing a rapid transformation. A new challenger, DeepSeek, has emerged from China, causing a stir in the industry with its advanced, cost-effective AI models and open-source approach. While American companies have long dominated the field, DeepSeek’s rapid advancements are prompting a serious re-evaluation of the future of AI dominance. Is this a sign of a shifting power balance, and what does it mean for the US AI industry? The situation is further complicated by an undercurrent of irony, as questions of data ethics swirl around the key players. So Is DeepSeek a threat to US AI?

DeepSeek: A New AI Powerhouse

Founded in 2023 by Liang Wenfeng, a graduate of Zhejiang University and co-founder of one of China’s top hedge funds, High-Flyer, DeepSeek is not your typical AI startup. Backed by High-Flyer’s resources and fueled by a mission to advance artificial general intelligence (AGI) through open-source research and development, DeepSeek has quickly made a name for itself. Interestingly, the company is primarily focused on research and has no detailed plans for immediate commercialization, setting it apart from many of its profit-driven counterparts.

DeepSeek’s arsenal features two key models that have caught the attention of the AI world:

• DeepSeek-V3: Released in late 2024, this model boasts a staggering 671 billion parameters and was trained on a massive dataset of 14.8 trillion tokens. Benchmark tests indicate that DeepSeek-V3 not only outperforms models like Llama 3.1 and Qwen 2.5 but also matches the capabilities of industry heavyweights like GPT-4o and Claude 3.5 Sonnet. All this, achieved at a training cost of around $5.58 million.
• DeepSeek-R1: Launched in January 2025, DeepSeek-R1 focuses on logical inference, mathematical reasoning, and real-time problem-solving. What makes this model particularly noteworthy is its training method – reinforcement learning without supervised fine-tuning, employing group relative policy optimization (GRPO) to enhance reasoning capabilities. This innovative approach has enabled DeepSeek-R1 to achieve performance comparable to OpenAI’s o1 across various tasks, including mathematics and coding.

DeepSeek-R1, in particular, has been described by some experts as “AI’s Sputnik moment,” suggesting it marks a turning point in the global AI race. On January 27, 2025, DeepSeek even surpassed ChatGPT as the most downloaded free app on the iOS App Store in the United States, a testament to its growing popularity and perceived capabilities.

The DeepSeek Threat: How It Challenges US AI

The DeepSeek threat to American AI is multifaceted and stems from several key advantages:

Cost Advantage

•  DeepSeek’s ability to achieve comparable performance to US models at a fraction of the cost is a game-changer. This cost-effective AI approach has the potential to disrupt the economic models of American AI companies, potentially triggering price wars and impacting profitability.
  • Investors are taking note, and this could lead to a reassessment of investment strategies and potentially lower valuations for some AI companies.
  • As stated in the Gemini Deep Research report: “By achieving comparable performance at a fraction of the cost, DeepSeek could trigger a price war in the AI industry, forcing companies to lower their prices and potentially impacting their profitability.”

Technological Competition

• DeepSeek’s success challenges the long-held assumption of US dominance in the AI field.
  • Despite US export controls on advanced chips, DeepSeek has demonstrated the ability to develop cutting-edge AI models utilizing older chips and innovative engineering practices.
  • This signals China’s growing technological capabilities and its determination to compete on a global scale.This could lead to a shift in the global balance of power in technology and innovation.

National Security Risks

•  DeepSeek’s ties to China have raised national security concerns in the US.
  • Some experts fear that the company’s AI models could be used by the Chinese government for purposes such as espionage or spreading misinformation.
  • These concerns are further amplified by DeepSeek’s data collection practices, which involve storing user data on servers in China.
  • “Some experts fear that the company’s AI models could be used by the Chinese government for purposes such as espionage or spreading misinformation,” highlights the Gemini Deep Research report.

Open-Source Disruption

•  DeepSeek’s commitment to open-source AI models poses a direct challenge to the proprietary models favored by many American companies.
  • Open-source models foster greater collaboration and innovation, potentially accelerating the pace of AI development.
  • This puts pressure on US companies to reconsider their strategies and potentially embrace more open practices to remain competitive.

Innovative Training Methods 

• DeepSeek relies heavily on reinforcement learning to develop reasoning skills in its models.
  • This approach differs from the hybrid training strategies employed by many US-based AI companies.
  • DeepSeek has less reliance on supervised fine-tuning and large labeled datasets.

Fragmentation of AI Development

•  DeepSeek’s success could lead to a fragmentation of AI development into three distinct paths: application developers, major labs, and domain experts.
  • Application developers might leverage open-source foundations like DeepSeek to create specialized AI tools for niche markets.
  • Major labs could continue to focus on general-purpose models with broader applications.
  • Domain experts might develop tailored AI solutions for specific industries using leaner architectures.

Transforming Energy and Infrastructure

•  DeepSeek’s efficient use of resources has implications for the energy and infrastructure sectors.
  • While DeepSeek’s models require less computing power, the increased adoption of AI driven by its cost-effectiveness could lead to a rise in overall energy and infrastructure demands.
  • This highlights the need for continued investment in sustainable energy solutions.

The Undercurrent of Irony: A Complex Ethical Landscape

The current AI landscape is not without its complexities. It is worth noting that OpenAI, a leading American AI company,has accused DeepSeek of using its data outputs, potentially through a process called distillation. Ironically, OpenAI itself has faced numerous lawsuits related to its own data scraping practices, as detailed in the Upper Echelon Gamers video “The DeepSeek Mania.” This situation highlights the ongoing debate surrounding data usage in AI development and raises questions about the ethical boundaries in this rapidly evolving field. The need for a broader discussion about AI ethics data scraping and the responsibilities of companies in this area is more pressing than ever.

Counterarguments and Perspectives

While the DeepSeek threat to American AI is a significant concern for many, some experts believe it’s overstated. For instance, Anthropic CEO Dario Amodei argues that DeepSeek’s V3 model, while impressive, is closer in performance to older US models and represents an expected point on an ongoing cost reduction curve. He suggests that US companies will likely respond by further optimizing their models and driving down costs. (short summary of video)

How American AI is Responding

American AI companies and policymakers are not standing idly by. Several strategies are being considered in response to the DeepSeek challenge:

• Increased Investment: Initiatives like the $500 billion Stargate project aim to boost US AI infrastructure and research, signaling a commitment to maintaining US competitiveness in AI.
• Regulatory Measures: Strengthening export controls, considering potential restrictions on DeepSeek’s operations,and formulating data security regulations are being explored to address national security concerns and limit DeepSeek’s access to US resources.
• Embracing Open Source: Some US companies are exploring the adoption of open-source AI models to foster collaboration and innovation, recognizing the potential benefits of this approach.
• Focus on Efficiency: There’s a growing emphasis on developing more efficient AI models that require fewer resources, mirroring DeepSeek’s cost-effective approach.
• Addressing Regulatory Barriers: Efforts are underway to streamline regulations and permitting processes for data centers and other AI infrastructure, aiming to accelerate AI development in the US.

Conclusion: DeepSeek a threat to US AI?

DeepSeek’s emergence has undoubtedly sent a ripple effect through the American AI industry. Its cost-effective and efficient models, coupled with its open-source approach, pose a significant challenge to the established players. While the long-term impact of DeepSeek remains uncertain, it has undeniably sparked a crucial conversation about the future of AI dominance and the strategies American companies need to adopt to maintain their leadership. The rise of DeepSeek might be a sign of a changing tide, highlighting the need for increased investment, innovation, and perhaps a re-evaluation of the balance between proprietary and open-source approaches in the race for AI supremacy. The coming years will undoubtedly be a defining period for the global AI landscape, and competition like this might be what is needed to push innovation even further.

Synthesis

The rise of DeepSeek as a potential threat to American AI companies is summarized in the following table:

Factor	DeepSeek’s Advantage	Potential Impact on US AI	US Response
Cost-Efficiency	DeepSeek’s models achieve comparable performance at a fraction of the cost of US models.	Could trigger price wars, potentially reducing the profitability of US AI companies and impacting investor confidence. This could lead to a reassessment of investment strategies and potentially lower valuations for some AI companies.	Increased investment in AI research and development, with initiatives like the $500 billion Stargate project aimed at expanding AI infrastructure. Focus on developing more efficient AI models that can achieve comparable performance with fewer resources.
Technological Advancements	DeepSeek has demonstrated the ability to develop cutting-edge AI models despite US export controls, utilizing older chips and innovative engineering practices.	Challenges US dominance in AI, potentially leading to a shift in the global balance of power in technology and innovation.	Strengthening export controls on chips and other technologies to limit DeepSeek’s access to advanced resources. Potential restrictions on DeepSeek’s operations in the US market.
Open-Source Approach	DeepSeek’s open-source models foster collaboration and innovation, allowing developers worldwide to access and modify its technology.	Creates pressure on US companies to reconsider their reliance on proprietary models and potentially embrace more open-source practices to remain competitive.	Consideration of open-source models by US companies to foster greater collaboration and innovation within the AI ecosystem.
National Security Concerns	DeepSeek’s ties to China raise concerns about potential use of its models for espionage, misinformation, and influence operations. DeepSeek’s data collection practices, including storing user data on servers in China, further amplify these concerns.	Increased scrutiny of DeepSeek’s operations by US policymakers and intelligence agencies. Potential regulatory measures to address national security risks, including restrictions on data collection and storage practices.	Regulatory measures to address national security concerns, such as stricter data security regulations and potential limitations on DeepSeek’s access to sensitive information.
Long-Term Implications	DeepSeek’s efficient and cost-effective approach, coupled with its open-source strategy, could reshape the global AI landscape.	Potential for increased competition from Chinese and other international AI companies, leading to a more fragmented and dynamic AI ecosystem. This could accelerate the pace of AI development and innovation globally, but also raise challenges related to intellectual property protection and international cooperation.	Continued investment in AI research and development, with a focus on efficiency and innovation. Fostering a competitive and dynamic AI ecosystem in the US through policies that support startups and encourage collaboration between academia and industry. Addressing national security concerns through a combination of regulatory measures and international cooperation.

Cited Works

1. What Is DeepSeek, the New Chinese OpenAI Rival?, retrieved February 1, 2025, https://time.com/7210296/chinese-ai-company-deepseek-stuns-american-ai-industry/
2. What is Chinese AI startup DeepSeek?, retrieved February 1, 2025, https://www.foxbusiness.com/markets/what-chinese-ai-startup-deepseek
3. DeepSeek vs. ChatGPT fuels debate over AI building blocks – VOA, retrieved February 1, 2025, https://www.voanews.com/a/deepseek-vs-chatgpt-fuels-debate-over-ai-building-blocks/7958031.html
4. What is DeepSeek? — everything to know | Tom’s Guide, retrieved February 1, 2025, https://www.tomsguide.com/ai/what-is-deepseek-everything-to-know
5. Chinese quant whiz built DeepSeek amid fund rout – The Times of India, retrieved February 1, 2025, https://timesofindia.indiatimes.com/business/international-business/chinese-quant-whiz-built-deepseek-amid-fund-rout/articleshow/117663090.cms
6. What DeepSeek Means for the Future of AI – Built In, retrieved February 1, 2025, https://builtin.com/artificial-intelligence/what-deepseek-means-for-tech
7. Is the DeepSeek Panic Overblown? | TIME, retrieved February 1, 2025, https://time.com/7211646/is-deepseek-panic-overblown/
8. 7 Implications of DeepSeek’s Victory Over American AI Companies – Medium, retrieved February 1, 2025, https://medium.com/@arijitghosh646/7-implications-of-deepseeks-victory-over-american-ai-companies-51610e6ac6c8
9. Q&A: What Is DeepSeek, the Bargain AI Roiling the Stock Market? – The Darden Report, retrieved February 1, 2025, https://news.darden.virginia.edu/2025/01/29/qa-what-is-deepseek-the-bargain-ai-roiling-the-stock-market/
10. Why DeepSeek is Sparking TikTok-Like National Security Fears | TIME, retrieved February 1, 2025, [https://time.com/7210875/deepseek-national-security-threat-tiktok/](https://time.com/7210875/deepseek-national-security-threat-tiktok/)
11. How DeepSeek changed the future of AI—and what that means for national security, retrieved February 1, 2025, https://www.defenseone.com/technology/2025/01/how-deepseek-changed-future-aiand-what-means-national-security/402594/
12. DeepSeek: A breakthrough moment for AI – KPMG International, retrieved February 1, 2025, [https://kpmg.com/us/en/media/

Summery of Upper Echelon’s Video

The YouTube video by “Upper Echelon” discusses the controversy surrounding OpenAI’s accusations against DeepSeek, an AI company, for allegedly stealing its data. The video highlights the irony of OpenAI, a company accused of scraping vast amounts of data without consent, complaining about DeepSeek’s alleged data theft. It recounts OpenAI’s history as a non-profit turning for-profit while aggressively raising funds and facing multiple lawsuits over data usage.

DeepSeek’s emergence with a cheaper, faster AI model trained with less computing power challenges OpenAI’s market dominance. The speaker argues that OpenAI’s complaints are hypocritical, given its own questionable data acquisition practices, and criticizes the tech industry’s lack of ethical consistency. The video suggests this situation exemplifies the concept of “stealing from thieves,” portraying both companies negatively.

(Summary made by NotebookLM)