Summary of Key Points
This article illustrates the transformation of the AI industry from a period of "conceptual frenzy" to a phase of "practical implementation," as reflected through 11 details from the World Artificial Intelligence Conference (WAIC): There are authoritative figures offering sober reminders about the "pseudo-intelligence" of large models, while companies are shifting their strategies from focusing on model parameters to generating real revenue. AI infrastructure (computing power and chips) has become a focal point at the conferences, with robot companies showcasing impressive displays while securing actual contracts. Cross-industry players (such as the emotional blogger Ququ) are also taking advantage of the trend, and the conference itself has turned into a platform for job hunting and networking. Overall, the AI boom has moved from a phase of storytelling to one of evaluating practical outcomes, with the entire industry seeking tangible ways to create real value.
1. Authority Figures Offering a Reality Check: AI Is Not Yet at the Stage of "True Intelligence"
On the opening day of WAIC, two leading AI experts dispelled the hype surrounding large models:
- Yuan Deng-Nian (Sutton), the father of reinforcement learning, stated that current large models do not possess true intelligence. He argued that merely accumulating computing power and imitating human data is insufficient; AI must learn by making mistakes in real-world contexts to discover new knowledge.
- David Bonio, a Turing Award winner, raised concerns about the governance of AI, warning last year about the potential for deception and self-protection in large models. This year, he questioned whether intelligence equates to power and who will control it.
The audience consisted of professionals eager not to miss out on the opportunities (FOMO), but these experts' remarks punctured the bubble of the belief that AI is already omnipotent. Current AI systems are still adept at imitation and are far from achieving true, independent thinking.
2. Large Model Companies Changing Course: From Competing on Parameters to Generating Revenue
In the past, companies competed based on the size of their model parameters and high performance scores. This year, they are focusing on practical outcomes:
- Kevin Lee (Li Kaifu) of ZeroOneAllThings shifted from developing models to creating "navigation maps" for the practical application of AI in businesses. He pointed out that the quality of China's top ten models is similar, and the real issue lies in the underutilized data within companies. His approach is to offer solutions that can improve financial reports, with the goal of generating revenue for both the company and its clients.
- Han Zhang (Yin Qi), a veteran from Megvii, is betting on the integration of AI into physical products (such as AI-driven car cockpits and smartphones). He has restructured his company into two divisions: "Brain" (Jieyue Xingchen) and "Body" (Qianli Technology, with Geely holding shares), and is pursuing an IPO in the Hong Kong stock market, aiming to succeed through real products and sales.
- Zhipu & Kimi were among the least hyped companies but remained low-key. Kimi released a new model with 2.8 trillion parameters before withdrawing from the conference early; Zhipu did not even participate. This suggests that they either have sufficient clients or are working on more significant initiatives.
In short, the era of storytelling is over; it's time to present tangible, profitable solutions.
3. AI Infrastructure Becomes a Focus: Servers and Chips Take Center Stage
This year, the highlight of the exhibitions was not model parameters but rows of computing servers and AI hardware:
- Computing Power Clusters: Huawei showcased a supernode with 1024 Ascend cards (56 times the scale of NVIDIA's solutions), while Alibaba, Baidu, and ZTE also displayed their own advanced systems. The reason is simple: Models are becoming increasingly complex (e.g., Kimi K3 with 2.8 trillion parameters), and a single chip can no longer handle the workload; thus, multiple cards are combined to form "supercomputers." Since domestic computing power lags behind NVIDIA, companies are compensating by increasing the number of cards used.
- AI-Smartphones: Companies like MiniMax, Nubia, and Jieyue Xingchen released AI-enabled smartphones, which can automatically perform tasks (e.g., booking flights and taxis) and remember user preferences. These phones represent the next frontier in AI applications, as they are widely used devices. Whoever captures this market first will gain a significant advantage.
The AI industry is transitioning from a focus on software to an emphasis on hardware infrastructure.
4. Robots: Showy Displays for Attraction, Practical Solutions for Business
The H3 hall's embodied intelligence section was the most popular, but behind the glamour lay the reality of the industry:
- Showy Displays: Robots that can sing, dance, serve tea, create sugar art, and even mimic Western girls were used to attract attention.
- Practical Solutions: The real moneymakers are the unassuming robotic arms. For example, four robots worked together on an automobile battery production line, automatically handling tasks without human intervention. Media joked that "the ones that can dance attract visitors, while the ones that can work generate revenue." This highlights that practical functionality is key in the robot industry.
5. Cross-Industry Participation: Emotional Bloggers Join the AI Trend, and Conferences Become Networking Platforms
WAIC is not just a technology exhibition but also a marketplace for resource exchange:
- Ququ: An emotional blogger who became famous for her search for a better partner and later faced tax issues, used her booth in a remote corner of the startup area to promote her new products. She explained that participating in the conference was a way to showcase her "tech image" to her private client base (with a membership fee of 199,800 yuan) and explore new revenue streams.
- Students Seeking Jobs: Many students brought their resumes to the conference, with companies like Mianbi Zhiheng listed as friendly for job applications. One graduate student noted that the experience was more valuable than obtaining a master's degree, as it provided opportunities to learn about the industry and make connections.
The AI trend is affecting all sectors, with both tech giants and cross-industry players seeking their share of the market.
Conclusion: A Critical Reflection on the AI Boom
The details from WAIC indicate that the AI industry is moving from superficial excitement to genuine progress. The cautious advice from experts, the strategic shifts of companies, the development of infrastructure, and the differentiation in robot technologies all point towards a focus on real value. However, the influx of cross-industry players also shows that the AI trend still attracts attention. Ultimately, only those companies that can solve practical problems and generate actual revenue will thrive in the long run. The future of AI lies not in empty concepts but in tangible actions and results.