US technocrats are struggling to compete with China’s Deepseek, and are seeking any reason to ban it
By ljdevon // 2025-01-30
Copy
 
As US technocrats and oligarchs realize they are losing the AI race with China, they are seeking any possible way to take down China’s more advanced, more logical, uncensored AI systems. US leaders are seeking to control the narrative. China’s Deepseek is not only outperforming US AI systems, but it's also more transparent with its reasoning and able to communicate more basic truths on contentious topics that are censored in the US. This week, top White House advisers expressed alarm over China's DeepSeek, a company that has allegedly leveraged a technique known as "distillation" to potentially benefit from U.S. AI advancements. The controversy raises critical questions about intellectual property, national security, and the future of AI development. As the U.S. contemplates stringent measures to curb China's AI progress, the situation underscores the complexities of regulating a field that thrives on open collaboration and rapid innovation – values that US technocrats are unwilling to behold as they are outsmarted by the Chinese.

US refusal to collaborate and the concocted "distillation dilemma"

At the heart of the issue is the distillation technique, a method by which one AI system learns from another. This process allows newer AI models to benefit from the extensive training and computational power invested in older, more established models. According to sources in Silicon Valley, this technique is not only common but also difficult to regulate or prevent. • DeepSeek's Innovation: DeepSeek recently unveiled an AI model that rivals the capabilities of U.S. giants like OpenAI, but at a significantly lower cost. The company also made the code freely available, further fueling concerns about potential misuse. • Terms of Service Violations: The distillation technique, while widely used, violates the terms of service of some prominent U.S. AI models, including those from OpenAI. OpenAI is reportedly investigating whether DeepSeek inappropriately distilled its models. • Industry Perspective: Naveen Rao, vice president of AI at Databricks, likened the practice to automakers examining one another's engines. "When it's extractable information, you're going to extract it and try to get a win," Rao said.

Regulatory challenges for power hungry US technocrats

The U.S. government's response to the situation highlights the difficulties of regulating AI technology. Howard Lutnick, President Donald Trump's nominee for Secretary of Commerce, vowed to impose restrictions on DeepSeek, citing concerns about misappropriation of U.S. technology. • Enforcement Difficulties: Technologists argue that blocking distillation is easier said than done. With open-source models like Meta's Llama and French startup Mistral's offerings freely available, detecting violations of terms of service is challenging. • Open-Source Accessibility: Umesh Padval, managing director at Thomvest Ventures, noted that "it's impossible to stop model distillation when you have open-source models like Mistral and Llama. They are available to everybody." • Potential Solutions: Some propose stringent know-your-customer requirements, similar to those in the financial sector, to prevent unauthorized use of AI models. However, such measures are not yet in place and may face resistance. The current situation echoes previous concerns about the semiconductor industry, where the U.S. has imposed restrictions on chip exports to China. The U.S. is also examining ways to restrict work on certain open technologies, reflecting a broader trend of tightening controls on technology transfer. The US is also concerned about the rise of TikTok, and wants to ban it too because they cannot control it, regulate and censor people like they do with US-based social media platforms. As the U.S. navigates the complexities of regulating AI technology, it must strike a delicate balance between protecting intellectual property and fostering innovation. The case of China's DeepSeek highlights the challenges of containing AI advancements in an interconnected world. While the U.S. seeks to maintain its competitive edge, it must also recognize the limitations of regulatory measures in a field that thrives on collaboration and rapid evolution. Sources include: Reuters.com Reuters.com ACLU.com
Copy