Last week’s release of the DeepSeek’s Chinese AI model, R1, featuring advanced open reasoning ignited widespread acclaim, with Silicon Valley’s AI tech leaders honoring the creation as an AI transformative achievement.
Despite US sanctions limiting access to advanced semiconductor technology, AI development in China, including DeepSeek’s resource-efficient method, continue to flourish. However, CEO of DeepSeek Liang Wenfeng admits that export restrictions remain an issue.
DeepSeek’s R1 Chinese AI model grabbed a lot of attention for its high performance and significant cots, at just $5.6 million. This is the opposite to the hundreds of millions spent by US companies, mainly OpenAI.
Reactions to DeepSeek’s breakthrough have been mixed, while venture capitalist Marc Andreessen called the achievement one of the best Chinese AI models.
Moreover, Y Combinator CEO Garry Tan said the low-cost Chinese AI model could benefit US companies by reducing training costs and increasing demand for AI.
However, Curai CEO Neal Khosla was suspicious that DeepSeek was challenging US competitors and suggested that the company’s low-cost claims were a deception to cause damage to US firms.
Community Note said that Khosla did not provide evidence to support his claim, and his father is also an investor in OpenAI.
US and Chinese AI Model Performance Comparison
Yann LeCun, Meta’s Chief AI Scientist, warned against viewing the Chinese AI model as a China vs US issue, emphasizing that DeepSeek’s success highlights the effect of open-source AI models. The use of open-source tools by DeepSeek such as PyTorch and Meta’s LLMA models highlight the need of open research in making AI Chinese model more advanced.
When it comes to performance, US AI models like OpenAI’s GPT series and Google Deep Mind’s models, have long dominated innovation, by leveraging massive investments, extensive computational resources, and access to large data sets.
The US model focuses on optimizing AI for a wide range of tasks, including natural language processing and more complex problem-solving capabilities.
In parallel, Chinese models are working to close the gap, similar to DeepSeek. Even though they boast very strong performance, these models are more cost-effective, avoiding the expensive, specialized hardware and cloud infrastructure used by US AI models.
Chinese Generative AI Model Affecting Markets
Since the release of DeepSeek’s AI assistant app, it has reached the top immediately on Apple App Store, even surpassing ChatGPT in downloads. China’s next generation AI plan focuses on providing alternative AI solutions and highlights the broader trend of open AI development, benefiting from shared research and worldwide collaboration.
Despite the US sanctions and the geopolitical tensions that are affecting Chinese AI models, the success of DeepSeek highlights that AI development have a potential to become more decentralized and competitive in the next years.
Alongside its affordable and advanced Chinese AI model and open-source approach, DeepSeek is challenging the basic AI development models and will definitely play a role in reforming the global AI industry.
Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Tech sections to stay informed and up-to-date with our daily articles.