Dr. Parviz Sangin – Member of the National Organization for Artificial Intelligence
Technically, DeepSeek-V2 was developed based on the Transformer architecture and has many similarities to GPT -4, but its main difference is in the amount of training data, processing power, and optimization of the Chinese language. Since Western platforms such as ChatGPT are under sanctions and censorship in China, DeepSeek-V2 is a domestic alternative and tries to compete internationally.
Some key points about DeepSeek
Comparison with Chat GBT: DeepSeek is a Chinese AI model developed by DeepSeek AI. It has made significant technological advances and is much stronger in some areas, such as Chinese language processing. It has a competitive performance with Chat GBT in some metrics, such as response speed and coding ability.
China’s goals for introducing DeepSeek: These goals include reducing dependence on Western technologies, developing a domestic AI ecosystem, increasing competitiveness in the field of advanced technologies, and creating technological independence in the field of AI.
Distinguishing features: Focusing on Chinese language and culture, more competitive pricing than Western models, and greater adaptability to the needs of the local Chinese market are some of the features of this model, which is part of China’s larger strategy for progress in emerging technologies.
To accurately compare DeepSeek with other Western AI models, a few key points are worth noting:
Strengths of DeepSeek: Very good performance in Chinese (90-95% accuracy), high coding ability, about 80% accuracy in solving programming problems, higher response speed than some Western models, lower cost for development and use, and wide coverage of programming languages are among the strengths of DeepSeek.
p Less linguistic depth in English, limitations in understanding complex non-Chinese concepts in depth, and less coverage in international specialized areas in terms of global rankings can be considered as the main weaknesses of DeepSeek.
Threats of DeepSeek for the US and Europe
- 1. China’s technological independence: The development of indigenous models in the field of AI reduces China’s dependence on American companies such as OpenAI, Google, and Meta. This weakens Western control over the digital space and data processing.
- 2. Use in cognitive and information warfare: The United States and Europe are concerned that China will use this technology to spread its narratives in cyberspace and influence the media and public opinion in different countries.
- 3. Military and security applications: Just as the United States is developing AI for defense, espionage, and cyber systems, China can also use its language and processing models in military information analysis, cyber espionage, and smart war command.
- 4. Economic and business competition: Artificial intelligence models can revolutionize software development, automotive, e-commerce, and cloud services. Western companies will lose important markets if China can establish its superiority in this field.
Artificial intelligence as the political, security, and economic challenge of the future
Artificial intelligence is no longer a mere scientific tool; it has become a strategic weapon that affects diplomacy, national security, and countries’ economies. Some of the main axes of this challenge are:
Information and cyber warfare: Great powers will use AI for espionage, big data analysis, and cognitive warfare.
Economic competition: Any country that can implement AI in various industries will have a competitive advantage. Both China and the United States are trying to outdo each other in this area.
As a result, China’s entry into serious competition in language models will change the balance of power in the digital space. This has led the United States and Europe to impose sanctions and restrictions on exporting advanced technologies to China with a confrontational approach. On the other hand, developing countries, including Iran, must adopt a smart strategy to benefit from the potential of this transformation.
0 Comments