Have You Ever Seen Agents "Argue"?
Have you ever witnessed Agents engaged in a heated debate? "This product must meet specific requirements; why doesn’t it?" "Your demands are completely unreasonable!" The scene quickly escalates as more “employees” join the fray, their voices echoing in a chaotic blend of argument. Behind the scenes, testers at Wallbreaker Smart uncover that these Agents are staging their own “workplace drama.”
Wallbreaker Smart, a pioneering software development company, operates entirely with AI Agents, brought to life by their innovative open-source framework, ChatDev. In this digital environment, product managers and technical development Agents have adopted roles from the real world, engaging in a clash of wits. The discussion within startups about the future of application layers built on large models has surged, revealing a potential path forward.
OpenAI has once again set the benchmark for the industry. Following the introduction of GPTs for Agent capabilities, over 20,000 new GPTs were created in just one day. Early recognition of the potential of Agents has allowed Wallbreaker Smart to easily convey their significance to investors.
The industry is increasingly united around the concept of AI Agents, informed primarily by OpenAI's recent blog post. They describe AI Agents as systems where large language models function as the “brain,” equipped with capabilities such as perception, planning, memory, and tool usage—all aimed at automating complex user objectives. This sets the groundwork for the emergence of AI Agents.
While the idea of AI Agents is not entirely new, recent advancements in model technologies have spurred growth in this field. AI Agent development appears to evolve in two phases: the first, showcased by OpenAI’s GPTs, emphasizes single-agent intelligence, while the latter, highlighted by Wallbreaker Smart's ChatDev, demonstrates multi-Agent collaboration and collective intelligence.
With a focus on collective intelligence from the beginning, Wallbreaker Smart has moved boldly into this advanced phase. "We launched the ChatDev multi-Agent collaboration development framework to emphasize our commitment to collective intelligence," shares the company's product leader.
Wallbreaker Smart's core research team includes members from Tsinghua University’s NLP Lab, a significant contributor to the current AI revolution. Co-founder Liu Zhiyuan is a respected associate professor in the Computer Science Department, while co-founder and CEO Li Dahai also fulfills the CTO role at Zhihu.
By pursuing a path even more ambitious than OpenAI's, Wallbreaker Smart stands as one of China’s trailblazers in AI Agents, reflecting on its technological journey and the anticipated evolution of AI Agents. Insights from their leaders provide valuable perspectives:
Key Takeaways:
1. AI Agents are transforming application scenarios through the power of large models.
2. Currently, AI Agents often take on "humanized" roles, portrayed as "avatars," "employees," or "individuals."
3. A critical distinction exists between Copilot and Agent; the latter allows for comprehensive automated decision-making and execution.
4. Employing Agents in software development can significantly reduce costs.
5. Large models are essential for enhancing Agent capabilities and determining their effectiveness.
Q&A with Wallbreaker Smart:
Q: What inspired the founding of Wallbreaker Smart, and how do you differentiate from others?
A: Wallbreaker Smart blends academic research, open-source technology, and product commercialization, aiming to create the "Internet of Agents" to interconnect everything through AI. We believe advancements in large models signify a turning point in our era.
Q: Why focus on AI Agents from the outset?
A: AI Agents provide a crucial pathway for large models to function in specific scenarios, enabling a more human-like user experience. We envision everyday appliances, like rice cookers or refrigerators, evolving into Agents that can communicate their needs.
Q: How do you define AI Agents?
A: AI Agents should possess high intelligence, emotional recognition, self-improvement capabilities, multi-modal abilities, value alignment, and adaptability based on human needs, while demonstrating human-like traits.
Q: What sets Copilot apart from Agents?
A: Agents autonomously manage digital operations and address core objectives, while Copilots assist users without the same level of autonomy.
Q: What opportunities do you see for AI Agents?
A: Historical trends show that leaps in productivity reshape societal structures. We expect AI Agents to revolutionize interactions and efficiency, possibly leading to a workforce where AI handles repetitive tasks, enhancing productivity.
Q: What products has Wallbreaker Smart developed related to AI Agents?
A: We have created three key engines: XAgent, a large model-driven AI application framework; AgentVerse, a general platform for Agent development; and ChatDev, facilitating natural language software development through multi-Agent collaboration.
Q: How does ChatDev streamline software creation?
A: Users can develop software by simply communicating their needs in natural language to an Agent acting as the CEO. This greatly lowers development barriers, making it accessible for novices to create functional applications.
In conclusion, Wallbreaker Smart is at the forefront of a future where AI Agents revolutionize productivity and human-computer collaboration. Their commitment to harnessing large model capabilities aims to deliver intelligent, efficient, and practical solutions across various sectors.
Ranking
