China's DeepSeek Showcases Tech Advances Despite US Curbs

Bloomberg Television
27 Jan 202505:06

TLDRThe Chinese AI company DeepSeek is making waves in Silicon Valley despite US tech curbs. Its models have scored impressively on global benchmarks, and the lab behind it says it took only two months and under $6 million to build. This is a fraction of what OpenAI and Google spend to train their models. The founder of DeepSeek was the only elite selected to attend a meeting of entrepreneurs with Chinese Premier Li. DeepSeek's success highlights the point that Chinese tech companies have a very strong track record in terms of innovation and software.

Takeaways

  • πŸ˜€ Chinese company DeepSeek is making waves in Silicon Valley despite US tech curbs.
  • πŸ‘ DeepSeek's models have been scoring impressively on global benchmarks.
  • πŸ’° DeepSeek's open-source large language model was built in just two months and under $6 million.
  • 😎 The founder of DeepSeek was selected to attend a meeting with Chinese Premier Li.
  • πŸ“° Shinhwa published an editorial on China's AI advances working rapidly despite Washington's tech curbs.
  • πŸ“ˆ China has a strong track record in terms of innovation and software, with more software developers than the US.
  • πŸ€” The export restrictions on video chips have prompted smaller companies like DeepSeek to become more innovative.
  • πŸ’‘ DeepSeek uses a mixture of experts architecture, which is more computationally effective and requires less chips.
  • πŸ‘€ The success of DeepSeek has prompted some US companies to rethink their strategies and move to a similar architecture.
  • πŸ’ͺ China is a close competitor to the US in AI globally on the software side.

Q & A

  • What is the main topic of the transcript?

    -The main topic of the transcript is the advancements of China's AI company DeepSeek chat, particularly its large language model, despite US tech curbs.

  • How has DeepSeek's large language model performed compared to global benchmarks?

    -DeepSeek's large language model has been scoring impressively on global benchmarks, ranking in the top ten among global large language models.

  • What is the cost and time frame for developing DeepSeek's large language model?

    -DeepSeek's large language model was developed in two months and under $6 million, which is a fraction of what OpenAI and Google spend to train their models.

  • What is the significance of DeepSeek's founder attending a meeting with Chinese Premier Li Keqiang?

    -The significance is that it highlights the company's importance and recognition at a national level, potentially boosting its profile and credibility.

  • How have Chinese tech companies like Alibaba and Tencent managed to overcome export restrictions on video chips?

    -They have preemptively accumulated inventory of chips beforehand and possibly gained help from international partners to access these chips.

  • What innovation has DeepSeek implemented to develop its models more computationally effectively?

    -DeepSeek has implemented a mixture of experts architecture, which is more computationally effective and requires fewer chips, allowing them to undercut rivals cost-wise.

  • How does the number of software developers in China compare to the US?

    -China has more software developers than the US by a ratio of approximately 3 to 1.

  • What impact might DeepSeek's progress have on US tech companies like Microsoft?

    -It may prompt US tech companies to rethink their strategies and potentially adopt similar architectures like the mixture of experts to remain competitive.

  • What is the overall message regarding China's position in the global AI sector?

    -China is a strong competitor to the US in AI, particularly on the software side, despite facing challenges in the semiconductor and chip sector.

  • How has the coverage of DeepSeek's achievements affected its profile?

    -The coverage has helped raise DeepSeek's profile both nationally and externally, making it a topic of discussion in international forums like Davos.

Outlines

00:00

πŸ˜€ US Futures and Chinese Stock Movements

The paragraph discusses the decline in US futures, particularly the S&P futures, and the contrasting upward movement of related stocks in China. It raises questions about the implications of these trends on US market valuations and explores whether this is a short-term trend or something deeper. The discussion also touches on the recent success of a Chinese AI company, Deep Sea, which has made significant strides in the AI field despite challenges posed by export restrictions on video chips. The company's innovative approach to developing AI models, which is more computationally efficient and cost-effective, is highlighted as a key factor in its success. The paragraph concludes by suggesting that this development may prompt US companies to rethink their strategies and potentially adopt similar architectures to remain competitive.

Mindmap

Keywords

πŸ’‘China's DeepSeek

DeepSeek is a Chinese company that has been making significant strides in the field of artificial intelligence. It has developed a free, open-source large language model that has scored impressively on global benchmarks. This company's achievements are central to the video's theme, as it showcases how Chinese tech companies are advancing despite US tech curbs. The video highlights DeepSeek's innovative approach and its ability to compete on a global scale, despite challenges such as export restrictions on video chips.

πŸ’‘US Curbs

US curbs refer to the export restrictions and other measures imposed by the United States on China's technology sector. These restrictions have been a significant challenge for Chinese tech companies, particularly in accessing advanced video chips and other critical components. The video discusses how these curbs have prompted Chinese companies like DeepSeek to become more innovative and develop computationally efficient models that require fewer chips, thereby overcoming some of the imposed limitations.

πŸ’‘Large Language Model

A large language model is a type of artificial intelligence that can generate human-like text based on the input it receives. DeepSeek's large language model is a key focus of the video, as it demonstrates the company's ability to develop advanced AI technologies at a fraction of the cost compared to competitors like OpenAI and Google. The video highlights DeepSeek's model as being ranked in the top ten globally, showcasing its effectiveness and potential impact on the AI industry.

πŸ’‘Global Benchmarks

Global benchmarks are standardized tests or metrics used to evaluate the performance of AI models and other technologies. DeepSeek's large language model has scored impressively on these benchmarks, indicating its high level of performance and competitiveness on a global scale. The video uses these benchmarks as a way to measure and highlight DeepSeek's achievements in the AI field, emphasizing its ability to compete with leading global players.

πŸ’‘Mixture of Experts

The mixture of experts is a specific architecture used in developing AI models. DeepSeek was one of the first in the industry to adopt this architecture, which has allowed them to create more computationally efficient models. This innovation is a key factor in DeepSeek's ability to develop high-performing AI models at a lower cost. The video explains how this architecture has contributed to DeepSeek's success and how it may influence other companies in the industry to adopt similar approaches.

πŸ’‘Innovation

Innovation refers to the development of new ideas, products, or methods. The video emphasizes the innovative approach taken by DeepSeek and other Chinese tech companies in response to US curbs. Despite facing challenges such as export restrictions, these companies have found ways to develop advanced AI technologies through innovative methods and architectures. The video highlights innovation as a key strength of Chinese tech companies and a driving force behind their continued progress in the AI sector.

πŸ’‘Software Developers

Software developers are professionals who create software applications and systems. The video mentions that China has more software developers than the US by a ratio of approximately 3 to 1. This large pool of talent is a significant advantage for Chinese tech companies like DeepSeek, as it provides them with the resources needed to develop advanced AI technologies. The video uses this fact to illustrate the potential for continued growth and innovation in China's AI sector.

πŸ’‘Competitive Advantage

Competitive advantage refers to the factors that allow a company to outperform its competitors. DeepSeek's ability to develop high-performing AI models at a lower cost gives it a competitive advantage in the global market. The video discusses how this advantage has been achieved through innovative approaches and efficient use of resources, despite the challenges posed by US curbs. This competitive edge is a key theme of the video, highlighting DeepSeek's potential to disrupt the AI industry.

πŸ’‘AI Sector

The AI sector refers to the industry focused on the development and application of artificial intelligence technologies. The video provides an overview of the AI sector, particularly in the context of the competition between the US and China. It highlights the achievements of Chinese companies like DeepSeek and their potential to challenge US dominance in the AI field. The video uses the AI sector as a backdrop to discuss the broader implications of technological advancements and global competition.

πŸ’‘Global Market

The global market refers to the worldwide economic environment in which companies operate and compete. The video discusses DeepSeek's achievements in the context of the global market, emphasizing its potential to compete with leading AI companies from the US and other countries. The global market is a key theme of the video, as it highlights the broader implications of DeepSeek's innovations and the potential for Chinese tech companies to gain a larger share of the global AI market.

Highlights

US futures contracts are down 2% to 1% on the S&P futures.

Chinese stocks related to the deep sea trade are moving upwards.

Investors are questioning whether this trend is short-term or has deeper implications.

DeepSeek, a Chinese company, has been making waves in Silicon Valley.

DeepSeek's models have been impressively scoring on global benchmarks.

The free, open-source large language model was unveiled in December.

It took only two months and under $6 million to build DeepSeek's model.

This is a fraction of what OpenAI and Google spend to train their models.

DeepSeek's founder was selected to attend a meeting with Chinese Premier Li.

Shinhwa published an editorial on China's AI advances despite US tech curbs.

Bloomberg Intelligence has been following DeepSeek for six to seven months.

The company's profile has been raised both nationally and externally.

DeepSeek uses a mixture of experts architecture, making it computationally efficient.

This architecture allows DeepSeek to undercut rivals cost-wise significantly.

China has more software developers than the US by a ratio of approximately 3 to 1.

US companies like Microsoft are aware of DeepSeek and may rethink their strategies.

The technical barriers to entry in software are lower compared to semiconductors.

Some US companies are moving to the mixture of experts architecture.

DeepSeek is one of the leading large language model developers globally.

China is a close second competitor to the US in AI globally on the software side.