AI Industry at a Turning Point as Chatbots Hit Their Limits, Executives Say
00:00/00:00
Listen to this article
1x
Subscribe now
Listen to the full version
Listen to the full version
Photo: AGI-Next
Several Chinese artificial intelligence (AI) executives warned that the chatbot-driven boom is nearing its limits and called for a pivot toward autonomous agents capable of handling complex real-world tasks, as they met at Tsinghua University to map out the industry’s direction for 2026.
At the AGI-Next Frontier summit on Saturday, speakers agreed that while domestic developers enjoyed a strong 2025 — particularly in open-source large language models (LLM) — chat-based interfaces are reaching their limits. They also warned that the technological gap between China’s models and elite U.S. closed-source systems remains wide and could even expand.
You've accessed an article available only to subscribers
VIEW OPTIONS
Unlock exclusive discounts with a Caixin group subscription — ideal for teams and organizations.
Subscribe to both Caixin Global and The Wall Street Journal — for the price of one.
- RELATED
Share this article
Open WeChat and scan the QR code
- DIGEST HUB
Digest Hub
Back
Explore the story in 30 seconds
- Chinese AI leaders warn the chatbot boom is nearing its limits, calling for a shift to AI agents able to handle complex real-world tasks.
- Despite significant gains in 2025, China still lags behind top U.S. closed-source AI models and risks widening the gap.
- Experts emphasize the need for multimodal, reliable, and context-aware AI, predicting future models will feature autonomy, personality, and enhanced safety alignment.
AI generated, for reference only
Who’s Who
- Zhipu AI
- Zhipu AI, founded by Tang Jie, a Tsinghua University professor, shifted its focus in 2025 to programming-oriented models and AI agents. It released the GLM-4.5 model in July and open-sourced AutoGLM (9B) in December. Zhipu AI believes the "chat" paradigm is reaching its limits and that future AI development should focus on multimodal integration.
- Moonshot AI
- Moonshot AI (月之暗面) is an AI company mentioned in the article. Its founder and CEO, Yang Zhilin, emphasized the focus on AI agents. In 2025, Moonshot AI launched Kimi K2, described as China's first agent model capable of complex multi-step tool execution. The company plans to further improve models' attention and reasoning for long, multi-step tasks.
- Meta Platforms Inc.
- Meta Platforms Inc. is mentioned as a US open-source rival whose Llama models were largely displaced by Chinese models on third-party leaderboards in 2025. This indicates strong competition in the AI market, with Chinese advancements making significant inroads against established US players.
- Apple Inc.
- Apple Inc. was mentioned by Moonshot AI founder and CEO Yang Zhilin, who quoted co-founder Steve Jobs to emphasize that scaling AI isn't solely about computing power, but also requires "taste" for true progress. This highlights the importance of distinct identities and "worldviews" in AI development.
- Tencent Holdings Ltd.
- Yao Shunyu, Tencent Holdings Ltd.'s chief AI scientist, highlighted a growing divide in AI applications between consumer and enterprise use. He stated that enterprise users prioritize reliability and precision, opting for stable, high-accuracy systems. Tencent plans to leverage its consumer strengths to develop models that better understand user context, focusing on individual situations rather than just increasing training scale.
- Anthropic
- Anthropic is a company that develops AI tools, including Claude Code. In 2025, tools like Claude Code demonstrated that businesses prioritize reliable, high-accuracy systems over cheaper alternatives requiring frequent human correction, highlighting a growing divide between consumer-facing and enterprise AI applications.
- Alibaba Group Holding Ltd.
- Lin Junyang, technical lead for Alibaba Group Holding Ltd.'s Qwen team, predicted that "AI training AI" will become a reality. He emphasized that AI must evolve from passive responses to actively interpreting environmental signals and taking action. Lin also warned about AI's current lack of understanding of human preferences and stressed the need for strong AI safety education.
AI generated, for reference only
What Happened When
- early 2025：
- DeepSeek is launched, causing a significant impact in the Chinese AI industry.
- 2025：
- Chinese open-source models occupy nearly all top five positions on the Artificial Analysis leaderboard, surpassing US open-source rivals.
- 2025：
- Moonshot AI launches Kimi K2, described as China's first agent model capable of executing hundreds of complex tool-invocation steps.
- July 2025：
- Zhipu releases its GLM-4.5 model.
- December 2025：
- Zhipu open-sources AutoGLM (9B), a lightweight, interaction-focused agent model.
AI generated, for reference only
- MOST POPULAR
Sign in with Google
Sign in with Facebook
Sign in with 财新