In recent months, the Chinese AI landscape has seen a surge in the development and release of AI Agent products, with nearly 50 different offerings now available. But with this growth, a crucial question arises: do these AI Agents possess the technical reliability to be effectively applied in various industries?
AI Agents: The Missing Link Between Models and Applications
According to the InfoQ Research Center’s China AGI Market Development Report, released in May 2024, AI Agents serve as the bridge between the model layer and the application layer, acting as a vital complement to large-scale model applications. In the past two quarters, the AI Agent field has undergone significant changes, prompting the InfoQ Research Center to analyze the current state of the technology, including technical frameworks, the gap between ideal and reality, and the backgrounds of manufacturers.
Technological Frameworks and the Brainpower of Large Models
Since the release of AutoGPT in March 2023, AI Agents have garnered attention for their autonomy and problem-solving capabilities. A variety of AI Agent technical frameworks have been developed, encompassing fields such as general-purpose, environmental simulation, software development, multimodal, translation, terminal interaction, and data analysis. However, the debate between single-agent and multi-agent systems continues.
The key question remains: is the brain of AI Agents, i.e., large models, sufficiently intelligent to support their practical application? Although large models have undergone three major updates and shifts in competitive focus within just two years, their performance in tool usage and real-world environment simulation still falls short. For instance, GPT-4’s success rate in the WebArena test was only 14.9%, and the newly released GPT-4o showed no significant improvement.
The Gap Between Ideal and Reality
Apart from the planning capabilities, there is a gap between the ideal AI Agent and the current state. The InfoQ Research Center has conducted an in-depth analysis of the differences between the ideal AI Agent and the current one in aspects such as autonomous thinking, tool usage, memory, and multimodal understanding. This situation has also raised higher requirements for the development and application of AI Agents, compelling technical teams to optimize system reliability to achieve comprehensive task execution capabilities.
AI Agents in the Chinese Market
In the Chinese market, numerous AI Agent products have been launched in various fields, providing technical solutions from different perspectives, such as workflow. The InfoQ Research Center has compiled a list of nearly 50 AI Agent products in the Chinese market, forming the China AI Agent Product Compass. The Compass will continue to be updated, and the InfoQ Research Center welcomes feedback and continuous attention from developers, readers, and manufacturers.
Manufacturer Backgrounds in the AI Agent Market
In addition to product reviews, the InfoQ Research Center has studied manufacturers that provide AI Agent services and found that their backgrounds mainly include large model startup companies, internet technology companies, RPA/process automation companies, and digital enterprise service providers.
Large model startup companies, such as Dify, Lannuo Technology, and Mirrorgate Intelligence, leverage their large model technology foundations to meet the practical application needs of enterprises. Their competitive advantage lies in their forward-looking perspective on large models and their provision of AI Agent application markets and development platforms for users.
Internet technology companies, such as Baidu, Huoshan Engine, and Tencent, offer complete AI technology solutions through their large models and AI cloud services. Their competitive advantage lies in their comprehensive AI ecosystem, including infrastructure, cloud, and large models. They have also launched a variety of large model-related applications and established a good user base and product iteration model.
RPA/process automation companies, such as Laiye Technology and Real Smart, integrate AI Agent technology into their existing RPA products, relying on their long-accumulated experience in enterprise process automation to provide more intelligent AI+RPA products and services.
Digital enterprise service providers, such as Youyou, Kingdee, Standard Cloud, and Dushi Technology, rely on their long-accumulated expertise in specific fields or industries to achieve functional upgrades for enterprise digital systems. For these manufacturers, AI Agents are often integrated into digital systems as functional components, achieving deep integration with existing digital systems through a comprehensive API linkage ecosystem, allowing users to experience AI Agents seamlessly.
Conclusion
The rapid development of AI Agents in China is a testament to the nation’s advancements in AI technology. While there is still much work to be done to ensure the reliability and effectiveness of these AI Agents, the progress made in the past two quarters is promising. With continued efforts from developers, researchers, and manufacturers, AI Agents are poised to become an integral part of various industries in the near future.
Views: 0