"The Economist" calls: 2025 is the era of "AI Agent", but we must pay attention to three difficulties

Here is the English translation of the text, with the specified terms preserved: The Arrival of the AI Agent Era Artificial intelligence (AI) is undoubtedly one of the most closely watched topics in the technology and crypto sectors today. Whether it's the sustained rise of US stocks last year or the wealth creation frenzy in the current crypto bull market, AI has played a crucial role. The AI Agent Era Has Arrived According to OpenAI's classification, the development of AI can be summarized into five stages: - Conversation: e.g., ChatGPT, capable of conversational chatting - Reason: like humans, capable of logical reasoning - Agent: capable of autonomous action - Innovation: capable of creative technological breakthroughs - Organization: the ultimate form, capable of managing and operating human organizations So, what stage are we currently in? The Economist reported last month that in the run-up to the Western Christmas, the AI sector had already entered a festive period - first, OpenAI launched its "12 days of shipmas" activity on December 5th, and successively released the latest products such as the generative tool Sora and the writing and coding tool Canvas; while another tech giant, Google, also quickly released the generative AI model Gemini 2.0, as well as the advanced products Astra and Mariner, announcing the official arrival of the AI Agent era. The Promising Future of AI Agents Compared to general chatbots, AI Agents are a type of AI that can take autonomous actions. They possess natural language processing capabilities and can also make decisions and interact. Regarding AI Agents, The Economist cited Alex Wang of the AI data company Scale, who stated: "Breaking through the chatbot stage could be one of the most important achievements in the tech industry in 2025, and this expectation has already driven up the stock prices of giants like Salesforce. The company said that its workplace AI Agents have reached agreements with over 200 clients within a week of launch, and even Microsoft, a powerful competitor, has also successively released numerous AI Agents." The Three Challenges in Developing AI Agents However, while The Economist's report shows an optimistic view and great expectations for AI Agents, it also states that developing AI Agents is more difficult than developing chatbots. Specifically, there are three reasons: 1. Data problem: Chatbots can answer questions by extracting information from the internet, while AI Agents need to process data on how to execute tasks, including sequences of operations and the underlying logic. In many cases, it is difficult to find sufficient data to train them. 2. Trust problem: Chatbots' flaws in their responses are easily exposed, but it is relatively more difficult to judge the correctness of an AI Agent's decisions. 3. Cost problem: To enable AI Agents to reason, plan, and solve problems, they require a large amount of computing resources, which also means a huge cost investment. AI Agents Ignite the Crypto Sector It is worth mentioning that since the launch of the AI meme coin GOAT and the listing of the AI meme coin ACT on Binance last year, the crypto sector has quickly sparked a wave of AI token issuance, and AI Agents have become one of the most closely watched topics in the crypto sector. However, observing the various AI Agents in the current crypto sector, unlike OpenAI's and Google's AI Agents, most of them are still wrapped in the shell of meme coins, mechanically pushing out news, and only a small portion have shown capabilities in analysis, reasoning, decision-making, and task execution. Regarding this, Haseeb Qureshi, a partner at the venture capital firm Dragonfire, recently commented that the current "AI Agents" craze sweeping the industry may be able to last until 2025, but will ultimately be unsustainable, and the real influential applications of AI will be in the fields of software engineering and security: "AI Agents will ultimately perish. These are not true agents, but just chatbots with memecoin wrappers. Apart from posting on Twitter, they have almost no agency." Further Reading: Will AI Be the Narrative Backbone of This Cycle? A Comparison of the Development Status of Various AI Agents Platforms