DeepSeek, a Chinese AI startup, has garnered international attention for creating cost-effective and efficient AI models. Yet, the U.S. government has started examining the potential “national security implications” of its use.
DeepSeek upon seeing the United States
The US National Security Council is thoroughly reviewing all information on DeepSeek, as stated by White House press secretary Karoline Leavitt.
This is a warning for the AI sector in America, and efforts are being made to protect the U.S. AI field.
David Sacks, the Tsar of White House artificial intelligence and cryptocurrencies, was asked in an interview about potential intellectual property theft in the growth of DeepSeek.
AI companies may start taking measures to hinder the practice of distillation, a technique in AI where a model learns from another model, which could potentially impede the development of imitating models in the future.
President Donald Trump recently mentioned that the Chinese app could motivate American companies and highlighted the positive aspect of a Chinese company developing a more affordable and quicker artificial intelligence technique.
The introduction of DeepSeek by a Chinese firm serves as a warning for our industries to prioritize competitiveness. Emphasizing the importance of being innovative and assertive in the market, this development can lead to cost savings and efficient solutions.
The US government is currently informally evaluating the DeepSeek, with potential consideration of a thorough code review and even a potential ban in the country.
The Rising of the DeepSeek
Established in 2023 by Liang Wenfeng, a former math prodigy and fund manager, DeepSeek has competed with tech giants like OpenAI and Google.
In December 2024, the company introduced “DeepSeek-V3”, a language model that competes with GPT-4o and Llama 3. Its standout feature is its capacity to offer top-notch answers and tackle intricate equations while using significantly less computational power and cost compared to other models.
The DeepSeek method combines open source methods and enhancements that enable more cost-effective training of AI models. For instance, training similar models in the US typically costs about $100 million, but DeepSeek achieved the same results for under $6 million by utilizing less powerful chips.
The company decided to release its open source model “DeepSeek-R1,” enabling global developers to use and alter the technology, fostering a spirit of cooperation in the field of AI.
Paraphrase: Origin