"DeepSeek didn't come out in a day, just as ChatGPT didn't emerge in a day two years ago," said Wu Zhiqiang, an academician of the Chinese Academy of Engineering who has long focused on and researched artificial intelligence. At an expert tour event held in Fuzhou, Fujian, on the evening of April 28, he revealed that there are at least 18 teams in China comparable to the level of DeepSeek, highlighting China's potential in the field of artificial intelligence.
It was noted that as early as 2014, the Chinese Academy of Engineering formed a professional team to comprehensively track global developments in artificial intelligence theory, models, technologies, and case studies every week, producing reports for the past ten years. "We have now completed about four to five hundred weeks of tracking. We started paying attention to ChatGPT three years before its release. Among the 16,000 AI teams we monitor, it received the highest score of 60 in our five-star rating system. It's just one among many."
Wu Zhiqiang stated that early large AI models absorbed various inputs during the "feeding" training phase, but often mixed genuine and false information. Additionally, the large-scale feeding led to enormous computational resource consumption. Since 2020, the Chinese scientific community has begun to shift its approach, focusing on breakthroughs with precise, specialized small models, using quality literature from various fields for training, carving out a distinctive path that achieves more accurate results with 5% to 10% of computational power.
Using Fujian as an example, Wu elaborated on the idea of developing artificial intelligence following the principle of "following nature." "Why is Fujian's ecology so good? Because it has sunlight, sea breezes, and mountain springs. To develop an AI ecosystem, we must learn from natural ecosystems—where the sunlight is, where the dew is, where the mountain springs are, and where the nutrient-rich soil is. If we understand these clearly, we can immediately connect the dots on how to quickly unleash and prosper an innovation ecosystem and AI ecosystem," he said.
Wu also emphasized the need to uplift and recruit more young people in the AI field, believing that, given time, they will accumulate knowledge and achieve great results. Three years ago, students from Wu's team developed an AI agent trained on papers he had written and published. "This 'AI Wu Zhiqiang' surprised me with its accuracy after I tested it for a long time," he noted, emphasizing that this agent was purely a product of his students' experimentation, without any formal assignment or planning. "Such 'youth emergence' is a crucial focus for China's future in the AI field; we must recognize, encourage, and nourish the youth, as they will undoubtedly create significant projects," Wu stated.
"I believe that any innovative team should bravely challenge the most difficult problems in their field. DeepSeek has solved many issues, such as the definition that only large computational power can produce high intelligence. Is it possible to achieve higher intelligence using less computational power? It has done so," said Liu Yunjie, an academician of the Chinese Academy of Engineering, expressing confidence in the development of AI in China. He believes that China has advantages in empowering industries with large industry models. "Our industries are the most comprehensive, and we have the most complete data, which is the most important foundation for AI. Combined with some open-source large models and good algorithms, I believe we can lead in the manufacturing sector," Liu added.
Chen Jun, another academician, emphasized the need to guide AI development correctly, stating, "We should delegate many strenuous and hazardous tasks to machines, while humans should become knowledge engineers who summarize and refine knowledge," thereby building a good AI ecosystem. "In the age of AI, many people are anxious, fearing that they or their children will lose their jobs. I believe that if we develop AI to the point of eliminating our professions and industries, that would be a grave mistake; we should not take this path," he said.
Related News:
Finance Spotlight | AI race sharpens competitive edge of China's national system
Comment