Raising lobsters and trading stocks: Is it "science" or "mysticism"?

Interface News Reporter | Liu Litong

Interface News Editor | Song Yejun

Recently, “Lobster Farming” (deploying, training, and using open-source AI agents OpenClaw) has exploded online, attracting many investors to join the wave.

Interface News has noticed that discussions about “using lobsters to trade stocks” on social media have become increasingly heated. Some marvel at how “lobsters” can monitor the market intelligently 24/7 with high efficiency and convenience, others lament that “using lobsters to trade stocks costs tokens that are more than 10 times higher than trading fees,” some are consulting about OpenClaw everywhere, and others are questioning the safety and reliability of “using lobsters to trade stocks”…

Since DeepSeek became popular last year, more and more A-share investors have begun to embrace AI in various ways, but their actual experiences vary widely.

Investor Chen Xue (pseudonym) has sought the “secret to wealth” across multiple AI large model platforms but has suffered an overall loss of nearly 20% in the bull market. She says, “All my sincerity was ultimately misplaced.”

Qin Peng (pseudonym), head of a quantitative team in South China, sees AI “partners” as a “trading research神器,” significantly boosting work efficiency.

What is the current performance of AI in stock trading scenarios?

Efficient, but not necessarily reliable

When encountering problems, the first reaction of Guangdong’s hot money trader He Feng (pseudonym) is to “check Baibao.”

Whether it’s breaking news or new thematic concepts, he can usually get a preliminary answer within 1-2 minutes. If deeper research is needed, he adjusts keywords and questions, and in just a few minutes, he can get a more satisfactory answer.

Before the emergence of large AI models, He Feng would spend a lot of time browsing news sites, stock forums, social media, etc., collecting enough information, then manually integrating and analyzing it to reach a somewhat acceptable conclusion.

Qin Peng prefers to combine his quantitative stock selection model with large AI models.

His quantitative model automatically screens stocks daily based on fund flow, market heat, price-volume trends, etc., then performs secondary filtering based on fundamentals and hot topics to identify final targets. With AI assistance, the time spent on manual filtering has shrunk from 3-5 hours to 30-50 minutes daily, greatly improving efficiency.

Additionally, when developing or modifying stock selection models, Qin Peng occasionally delegates simple tasks to AI.

“Efficiency” is the first keyword many investors think of when discussing AI stock trading. There are over 5,000 listed companies in A-shares, with continuous 24-hour updates of various financial information. Extracting the needed parts from this massive data set far exceeds any individual investor’s capacity, and AI finds this “small dish.”

However, many interviewees also agree that AI large models often provide unreliable answers.

For example, asking an AI model about the relationship between a stock and a hot topic usually yields a seemingly well-reasoned answer, but much of the content lacks factual basis.

Some investors cite examples where AI is asked to find the 10 stocks with the lowest PE ratios in the market. It only pulls data from dozens of stocks and provides an answer, some of which may be outdated or even incorrect.

AI “partners” also often display a “people-pleasing personality.”

For instance, if you ask, “Is A better than B?” it will list many supporting points. But if you reverse the question to “Is B better than A?” it will do the same. If you first ask it to analyze a certain industry, then ask which industries are worth watching now, the previously mentioned industries often appear again.

Almost all interviewees have experienced “AI hallucinations” during stock trading, where AI’s answers seem reasonable and comprehensive but are actually filled with fabricated facts, data, events, or even violate basic common sense—essentially “nonsense” in a serious tone.

In investing, any decision mistake can lead to real financial losses. These phenomena cause additional problems: although investors can get an answer from AI in minutes, they often spend multiple times that correcting or adjusting the AI’s output or repeatedly rephrasing questions to get more reliable responses.

Where are the problems?

Chen Xue first decided to try AI large models for stock trading after learning that DeepSeek’s underlying quantum algorithms are impressive.

Many top quantitative private funds publicly state they are deploying AI, but few outsiders truly understand what role AI plays in their investment decisions or how much it impacts their returns—how much of it is contributed by high-frequency trading.

A person from a leading Shanghai private fund believes that asking AI casually during stock trading is a completely different concept from applying AI in quantitative investment.

Generally, quantitative investment uses mathematical models, statistical methods, and computer programs to replace subjective judgment, characterized by discipline, data-driven decisions, diversified holdings, and strict risk control.

For most ordinary investors using AI models, the core decision-making is still human-driven, falling into subjective investment. Their holdings are usually limited, making it difficult to hedge risks caused by AI decision errors through diversification.

Furthermore, many investors are accustomed to using general large models like Doubao, Qianwen, DeepSeek, etc., which differ fundamentally from proprietary AI models developed by private funds.

According to Interface News, private funds mainly focus on three AI elements: data, computing power, and algorithms.

An industry insider told Interface News that data is the foundation for training AI models. In finance, high-quality, real-time, complete data is especially critical. General large models are mostly trained on text data, lacking sufficient high-quality financial data.

In terms of computing power, although general large models may involve more hardware investment than individual private funds, their broader scope and larger training datasets mean higher overall resource consumption.

On the algorithm front, leading private funds generally adopt a “self-developed” approach. Their underlying algorithms are often similar to those of general large models, but their fine-tuning directions and core algorithms are usually top secrets, rarely disclosed.

This insider also mentioned that some brokerages and institutions are actively promoting vertical AI models tailored for finance. While these organizations focus more on finance and have access to the latest financial data, their AI research is largely limited by computing costs and compliance constraints, making it difficult to fully meet investors’ expectations.

“Even if their AI models differ greatly from general large models, these models contain a wealth of investment knowledge. Why can’t they provide more reasonable investment advice like those top subjective investors?” many investors, including Chen Xue, have wondered.

In response, Chengdu hot money trader Ren Yu told Interface News, “Subjective investors may not demand the same level of data precision as quantitative investors, but their decisions still need to be based on the latest, relatively accurate data. General large models often fetch data that isn’t timely and may include some contaminated information, making their analysis unreliable.”

“The key issue is that AI large models lack a complete investment system. Every investment strategy has its characteristics and suitable market environment. From different strategic perspectives, conclusions about buy or sell points can be entirely different. For example, a stock might be a good buy from a medium- to long-term perspective but should be sold in ultra-short-term trading. AI models have learned many strategies but lack practical data supporting these strategies, making it hard to distinguish the true logic behind them,” Ren Yu explained.

Would feeding AI with the investment frameworks and philosophies of top investors improve its recommendations?

Qin Peng, who has tried this, opposes the idea. He believes that the “input” consists only of publicly available views and logic, but top investors often do not or cannot fully disclose their investment philosophies. Moreover, their investment systems evolve with market conditions.

Even if AI could give more reasonable investment advice, would investors strictly follow its strategies? The answer is probably no.

Human-AI collaboration is the consensus

How should ordinary investors use the efficient but potentially unreliable AI “partners”?

“Relying solely on AI for investment decisions is definitely not feasible. You still need to establish your own investment system,” Chen Xue concluded after over a year of experience.

Recently, she has paused real trading, deciding to focus on learning more investment knowledge. Once she is more skilled, she plans to restart her trading. During this process, she also discovered a new bright spot of AI models: “Their text analysis ability is really impressive—searching and summarizing investment knowledge is excellent!”

Qin Peng, who is relatively satisfied with AI “partners,” shared his experience with Interface News. He said that during the information collection phase, AI’s efficiency far surpasses humans, so this part can be delegated more; in the analysis phase, AI is also more efficient but prone to errors, so adjusting questions and prompts can help improve analysis; the decision-making stage is more complex and critical, requiring more subjective judgment from investors.

Most interviewees believe that in the future, AI large models will become more user-friendly, and many specialized AI models for finance and investment will emerge. However, AI will not completely replace human decision-making but will serve more as an auxiliary tool. Human-AI collaboration will remain the trend.

On one hand, AI models are trained by humans; how much computing power is invested, what data is fed, and which algorithms are used are all decided by humans. For the foreseeable future, AI cannot fully operate independently without human control.

AI models generally find patterns in historical data, but stock markets never repeat exactly. “Black swan” events can happen at any time, and AI inherently lacks the ability to handle such scenarios. Therefore, a truly all-powerful “agent” in investing is unlikely.

On the other hand, from a technical perspective, AI trading might someday outperform humans. But a series of potential risks makes it almost impossible to entrust all decisions to AI.

For example, strategy convergence has long been a concern in investing. As more institutions and investors use similar data and methods to train AI, strategies tend to become homogeneous, increasing correlated trading and amplifying market volatility, which could trigger systemic risks.

Moreover, AI models often operate as “black boxes,” making their decision processes difficult to trace. In case of anomalies, the models cannot be held responsible, and it’s hard to determine whether human factors influenced the outcome. If decision-making is fully delegated to AI, some groups could manipulate or influence the AI to control the market more covertly. From a regulatory standpoint, to prevent such risks, the application of AI in finance will likely be limited within certain boundaries.