.png)
November 29, 2024
Development in the field of Artificial Intelligence is progressing rapidly. While US companies like OpenAI, Google, and Meta have long held the leading role, Chinese AI models are increasingly entering the market and achieving considerable success. In particular, the models from Alibaba (Qwen) and DeepSeek (R1) are attracting attention and demonstrating that China is a serious competitor in the global AI race.
DeepSeek, a Chinese company operating in the field of quantitative hedge funds and AI investments, has introduced DeepSeek R1, an AI model that is in direct competition with OpenAI's o1. Similar to o1, DeepSeek R1 relies on improved reasoning capabilities to solve complex problems in areas such as science, programming, and mathematics. Initial benchmarks show that DeepSeek R1 can keep pace with OpenAI o1 in tests like AIME and MATH and surpass models like GPT-4o and Claude 3.5 Sonnet.
Despite the promising results, DeepSeek R1 is not without flaws. Examples show that the model can make suboptimal decisions in strategic games like Tic-Tac-Toe and has difficulties with complex logic puzzles. The avoidance of political questions regarding China also suggests a deliberate restriction by the developers to avoid potential conflicts with the Chinese government.
Alibaba has released Qwen, a family of AI models characterized by open source and multilingualism. Qwen is trained in 29 languages and thus addresses a significantly broader user base than many Western models, which are primarily focused on English. The open-source strategy promotes further development of the model by the community and enables faster adaptation to various use cases.
Qwen also shows strong performance in benchmarks, especially in the areas of logical reasoning and multilingual understanding. With QwQ-32B-Preview, Alibaba has also introduced a model specifically designed for analytical reasoning and complex tasks in mathematics, science, and programming.
The success of DeepSeek R1 and Alibaba Qwen is no coincidence, but rather embedded in a comprehensive AI strategy of China. Massive investments in research and development, access to huge amounts of data, and a growing pool of AI talent contribute to the rapid progress. The focus on the practical application of AI in various sectors underlines China's ambitions to take a leading role in the field of Artificial Intelligence.
The open-source strategy of Alibaba and the increasing availability of powerful AI models on platforms like Hugging Face are changing the dynamics in the AI field. The community plays an increasingly important role in the development and improvement of AI models and contributes to the democratization of access to this technology.
The competition between Chinese and Western AI models will further accelerate progress in Artificial Intelligence. The different strengths of the models – open source vs. closed source, multilingualism vs. specialization – offer users a growing selection of tools for various use cases. It remains exciting to see how the global AI race will develop in the future.
Bibliographie: https://www.heise.de/en/news/DeepSeek-R1-New-AI-language-model-with-reasoning-from-China-against-OpenAI-o1-10082115.html https://www.linkedin.com/posts/niels-rogge-a3b7a3127_its-another-crazy-week-for-open-ai-so-heres-activity-7220050101656764418-G952 https://medium.com/thedeephub/the-shifting-sands-of-ai-how-alibabas-qwen-signals-china-s-rise-in-the-global-llm-a02346ad1c6a https://github.com/nielsrogge https://www.youtube.com/watch?v=UmLuN14O0Vc https://restofworld.org/2024/alibaba-qwen-ai-model/ https://www.reddit.com/r/OpenAI/comments/1gvkjib/chinese_ai_startup_stepfun_up_near_the_top_on/ https://www.youtube.com/watch?v=IszSjZEvZZM