Editor's note: Recently, the open-source AI intelligent agent execution framework OpenClaw has sparked a "lobster farming craze." This year, references to "intelligent agents" were included in the government work report for the first time, signaling a potential new stage of development in the artificial intelligence industry.
【Anchor】Recently, the open-source AI intelligent agent execution framework OpenClaw has sparked a "lobster farming craze." This year, references to "intelligent agents" were included in the government work report for the first time, signaling a potential new stage of development in the artificial intelligence industry. Welcome to DDN Business Insider. Today, we specially invite Xia Chun, the founder of U&I Group and vice president of the Hong Kong Institution of International Finance, along with Jiang Han, senior researcher at the Pangoal Institution, to discuss industry developments and market opportunities in light of the AI intelligent agent boom. Hello everyone!
OpenClaw, nicknamed "lobster" by many netizens due to its red lobster logo, has recently attracted considerable attention. Mr. Jiang, could you explain the key differences between large models and intelligent agents? Does this evolution reflect a new trend in the field of artificial intelligence?
【Jiang】We see that there is a fundamental logical distinction between large models and intelligent agents. From an operational logic perspective, large models focus on cognition and generation; at their core, they rely on probabilistic text prediction and content creation, serving as passive question-and-answer tools. In contrast, intelligent agents, much like the "little lobster," center on perception and execution, possess the ability for autonomous planning, tool invocation, and interaction with the environment. They can transform abstract instructions into concrete multi-software operations, realizing a paradigm shift from dialogue to actionable tasks. Moreover, this transition indicates that artificial intelligence is moving from technical showmanship to tangible productivity; AI is no longer just an auxiliary for information retrieval but is becoming a digital employee capable of independently completing tasks. This marks the entry of the industry into a new stage of the intelligent agent economy, where the focus of competition shifts from the sheer scale of model parameters to ecological integration capabilities and scenario penetration abilities. The open-source community has become the core engine driving technological inclusivity and application explosion, foretelling that AI will deeply participate in and reshape the operational processes of the real economy.
【Anchor】Great. Then, Mr. Xia, what do you believe are the differences between large models and intelligent agents in terms of operational logic? What new trends do you think these differences reflect in the field of artificial intelligence?
【Xia】Regarding intelligent agents, especially the "little lobster," they can execute tasks from perception to planning to execution and constantly iterate. Therefore, they possess decision-making and action capabilities. They can help you check the weather, book flights, control your devices, assist with office tasks, and accomplish complex missions through interaction with the real world. They essentially have both a brain and hands, which I think is the biggest distinction. The overall development process shows that artificial intelligence is transitioning from passive-generating functionality to autonomous execution and evolving from a technical concept into a strategy and industry initiative that countries wish to promote. I see this as a significant trend.
【Anchor】This year, the government work report for the first time mentioned "creating a new form of intelligent economy." Mr. Jiang, what policy directions and development signals do you think this conveys?
【Jiang】The inclusion of intelligent agents in the government work report is significant. First, it signals that the state views intelligent agents as a key mechanism for cultivating a new form of productive forces. This marks a transition in the artificial intelligence strategy from basic large model development to large-scale industrial applications, shifting the policy focus from building models to using intelligence.
Second, the report explicitly mentions ways to create a new form of intelligent economy, meaning intelligent agents will be incorporated into the realm of new infrastructure construction. Through supporting policies such as computational collaboration and data flow, institutional safeguards are established for intelligent agents in vertical fields like government services, manufacturing, and finance. Furthermore, by seizing the global high ground in the intelligent economy and encouraging open-source innovation and industry extension, a self-controlled intelligent agent industrial chain is being constructed. This can effectively address international technological competition and promote a systematic leap for China's economy from digitization to intelligence.
【Anchor】Great. Mr. Xia, considering the current stage of artificial intelligence development, what important implications do you think arise from the inclusion of "intelligent agents" in the government work report for the first time?
【Xia】The core signal is that AI has shifted from a race focused on large models to competition centered on industrial implementation. The state has defined intelligent agents as the core mechanism for AI industrialization, aiming to create a new form of intelligent economy through them. Previously, it was a technical concept, but now it has transformed into a national strategy.
This means that intelligent agents have become a vehicle for AI development. While the emphasis was initially on large models for language generation, intelligent agents can now execute tasks autonomously and drive missions, helping industries take root. This time, intelligent agents and intelligent terminals are paired together, with the hope of forming a dual-drive system with terminals and intelligent agents that can create a complete application ecosystem.
【Anchor】Yes. However, amidst the "lobster farming" craze sweeping the market, the Ministry of Industry and Information Technology has issued relevant risk warnings. Mr. Xia, what do you believe are the core risk points currently concentrated on the application of this open-source AI intelligent agent?
【Xia】The primary concerns correspond to the following points: First, there may be a lack of strict review for plugins, which could lead to the implantation of malicious code that steals passwords or creates backdoors. This is the first corresponding risk. Second, large language models may exhibit deviations in understanding, resulting in actions such as deleting files or altering core data.
Another risk is that permissions may become uncontrollable, leading to information leaks. Elevated permissions might allow certain accounts to be compromised, resulting in the exposure of personal privacy and business secrets. Lastly, the internal networks of enterprises could allow horizontal spread and cross-system access, creating gaps in compliance that raise legal risks. Thus, while the autonomous execution of intelligent agents represents their value and capabilities, it also serves as a source of risk.
My personal advice is that during the process of using intelligent agents for "lobster farming," attention should be paid to several points. First, when using the "little lobster," it should be managed on an independent intelligent terminal with appropriate physical isolation. Additionally, authorization should adhere to the principle of least privilege, especially for operations involving deletions or modifications, which must require manual second confirmation.
【Anchor】Mr. Jiang, what do you believe are the core risk points regarding "lobster farming"?
【Jiang】First and foremost, the most significant hidden danger is permission control going awry. OpenClaw may need to take control of the highest permissions for file reading and writing, and software operations on users' devices. If default configurations are handled improperly or maliciously altered, sensitive data leaks could easily occur, or devices might be remotely controlled, leading to criminal risks associated with the illegal acquisition of computer information data.
Secondly, the ambiguity of trust boundaries might pose an abuse crisis. Intelligent agents have autonomous decision-making capabilities, but if effective ethical constraints are lacking, they could be used for automated cyberattacks, spam messaging, or financial fraud, such as automatic tipping or malicious ordering, and liabilities for these actions can be difficult to determine.
Thirdly, the prevalence of numerous third-party services can lead to serious chaos. Due to deployment barriers, the market has seen a surge of paid imitation services, which often mix in black industry chains that implant backdoors or steal keys. Users who enjoy convenience can unknowingly become victims or stepping stones for cybercrime, with security risks being exponentially magnified.
【Anchor】With the rapid rise of "lobster farming," related concept stocks in the capital market have already experienced a surge. Previously, the emergence of DeepSeek also drove up enthusiasm in relevant sectors. Mr. Xia, do you think Lobster AI will replicate the stock market effect that DeepSeek had at that time?
【Xia】It can be said that there are already some stock market effects, but they may differ in scale from those triggered by DeepSeek. The market's response is likely to be more structural and characterized by differentiation, with certain impulses still needing to be observed and possibly being relatively short-lived, not reaching the level of a complete sector-wide rally like DeepSeek.
At that time, when DeepSeek emerged, the whole A-share and Hong Kong stock markets were still in a relatively sluggish state. The emergence of Lobster, on the other hand, is more of a change at the application level. It is not a revolutionary model shift, but rather an application leap. Of course, the benefiting chains remain lengthy, decentralized, and somewhat unclear. Thus, there are greater opportunities on a local scale, but a lack of sector-wide targets.
【Anchor】Mr. Jiang, considering the current capital market's attention towards the AI sector, do you think Lobster AI can replicate the stock market effect seen during the emergence of DeepSeek?
【Jiang】To be honest, there are similarities in that both follow a speculative model triggered by the resonance of technological accumulation and policy support. Both benefit from the elevation of national strategies regarding artificial intelligence and, through open-source models, reduce usage thresholds, creating widespread attention and traffic effects, which have resulted in a strong herd effect in the capital market.
However, the differences are also obvious. The underlying logic of DeepSeek's market was based on computational power and model capabilities, benefiting upstream chip manufacturers, servers, and self-developed large model firms, falling under the investment logic of hard technology. In contrast, the underlying logic of the Lobster AI market is application scenarios and traffic distribution, benefiting downstream software integration, end-user devices, and security service providers, falling under the logic of business model innovation.
Moreover, the structure of risk appetite is different. The large model market is primarily driven by institutional funds, which focus on long-term technological barriers. In contrast, the intelligent agent market is more susceptible to retail and speculative sentiments, emphasizing short-term event catalysts and narratives, leading to greater volatility and stronger speculative characteristics.
【Anchor】Given this, Mr. Jiang, what changes do you anticipate in the overall focus of speculation in the AI sector in the coming year? How will developments on the industrial side guide the capital market's focus?
【Jiang】The focus of speculation will change. Firstly, it will shift from cloud-based large models to edge intelligent agents. The market's attention will no longer be on whose model parameters are larger, but on who can successfully deploy AI on terminals like smartphones, computers, and robots for true localized autonomous execution, making edge chips and operating system manufacturers the new favorites in the market.
Secondly, the investment logic will change from technical supply to scenario demand. Companies that solely tell the story of algorithms will become marginalized, while vertical application companies with specific industry data that can solve pain points and achieve commercial viability will receive higher valuation premiums.
Thirdly, the security and governance sector will become increasingly important. As intelligent agents penetrate deeper into social operations, security service providers focusing on permission management, data privacy, and ethical compliance will move from the periphery to the center.
【Anchor】OK, thank you to all the guests. That's all for this episode. Remember to follow us on YouTube or download our APP. I'm Yunfei Zhang, thanks for watching, and see you next time.
Anchor: Laura Cheung | Edited: Kelly Yang, Laura Cheung, Rachel Liu | Translate: Kato Ip | Proofread: Chris Liu
Related News:
DDN Business Insider | Oil prices hit record high: Why has HK become 'haven' for capital?
Comment