AI applications are in full swing, and all kinds of experience sharing have sprung up in all corners of the Internet.
One of them is to teach you how to use large models to do fund investment - from the most basic knowledge summary to reading quarterly reports, to performance data collation, premium rate analysis, etc., anyway, all kinds of imaginary needs, I have seen relevant introductions.
Is AI a "good partner for raising the base" reliable? I have also communicated with some people about this problem, and my attitude is to at least use a general-purpose AI assistant to assist in the investment of the base, which is to some extent self-inflicted.
An AI assistant that can only read web pages is not enough
Using AI assistants to assist fund research, in fact, when ChatGPT was just hot, there were many early adopters, but they all returned disappointed.
In fact, it is not surprising that the questions that ChatGPT was able to answer at that time were up to the data entered at the time of training, so it could not answer events that occurred after April 2023, which is undoubtedly a great limitation for the rapidly changing investment market. Not to mention that the big model has to have an "illusion" from time to time to introduce you to some funds that do not exist.
The reason why various AI-assisted investment funds have become a hot topic again recently is that more and more large models provide the function of "online search". By accessing the latest web materials, the large model has the possibility of keeping pace with the times, and it seems that the usability has improved a lot.
However, judging from my trial experience of several general-purpose AI that can be searched online in China, this kind of function is still far from being easy to use.
The chart below is an example of a test I did the weekend after the close of trading on April 19th. Just looking at the first few articles, you can know the latest net value, the increase in different time periods in the past, etc., and it really looks like it. But if you look closely, there are quite a few problems. One, I'm asking about how I've fared this year, and I'm really wondering about year-to-date (YTD) yields, which are an important indicator, but unfortunately the answers don't have the information I really need, and second, the answers are mixed in with a lot of "old news" from 2023, which is causing distraction.
This is actually a good result after I tested a few general-purpose AI, and some even returned data from April 17, and I can't even guarantee the timeliness of networking.
Of course, I am not surprised by this outcome. As an investor who has some understanding of large models, I have always felt that using general-purpose AI, even using online search to engage in fund investment, is essentially a wrong path. And industry vertical AI, which combines built-in raw data and professional content, may be a good way out.
Outside of searching, AI that understands you is a good assistant
Why is the networked search function of general AI not very helpful for fund investment?
To a large extent, it is still a matter of web pages as a source of information, and the current network information is uneven, and the answers of general-purpose AI assistants largely depend on the quality of the searched web pages.
In contrast, I asked Zhi Xiaobao, Alipay's built-in AI financial assistant, the same question, and its answer was not only public market data, but also integrated the latest views of a large number of professional institutions in the Ant Fund ecosystem. For example, if you ask "can you still buy gold funds now?", the overall medium- and long-term optimism of the institution is clear at a glance, and of course, there are many risk warnings after the short-term rise. In terms of information increment, Zhi Xiaobao's answer has a lot of dry goods.
Of course, the induction and summary of ready-made information is essentially the processing of text. However, in terms of fund investment, the more important thing is the linkage of unstructured data.
For example, when I express that I want to buy a certain fund, the general-purpose AI assistant will at most search the whole network and summarize the characteristics of the fund that others have written. It can't be said that these contents are worthless, but when it comes to investment and financial management, it is not enough to consider the advantages and disadvantages of a single product, but to consider it comprehensively under your financial needs and existing financial allocation.
When I asked Zhi Xiaobao if he wanted to buy "Southern Value Selection", Zhi Xiaobao would first substitute my position record and tell me what I had held and what the income was. On this basis, it will also combine my overall financial holdings to provide more in-depth analysis and reference conclusions.
I tried the short-term bond fund of Tianhong Amway again, and Zhi Xiaobao will call the professional allocation tool of three money, combined with the current situation of my stable assets, mainly money market funds, to remind me to pay attention to the structure of stable assets and pay attention to fixed income assets.
Vertical AI applications are available, but work is still needed
Connecting the underlying data can give AI applications in vertical industries a starting line. Of course, the advantages of these financial vertical AI applications are not only in opening up the underlying data.
More importantly, in terms of financial service functions, these institutions are more professional, but they are too rich and complex, and they are not very friendly to ordinary people.
However, the emergence of large models provides a new possibility - I think that in essence, large models are a new UI that interacts with text in a more friendly way to reach the original function points of each family.
Here's an example, when I asked Zhi Xiaobao in Alipay to "recommend several active funds with positive returns in the past year in the large-cap value style", Zhi Xiaobao directly called Ant Fund's original "conditional base selection" tool to give references with "value mining" and "high returns in the past 1 year".
I've always believed that the LLM model is a new UI, and that's exactly what it is.
In fact, at last year's Bund conference, I had an exchange with the product manager of Zhi Xiaobao, and at that time I was concerned about how to avoid the "illusion" problem that everyone was afraid of at that time after Zhi Xiaobao was empowered by the large model, and the product manager said that the financial scenario has extremely high requirements for accuracy, so the core data, decision-making, etc. will call special functions, and the task of AI is to understand the corresponding special functions called by users.
In fact, the function of conditional selection to screen the structured data of the fund is not difficult to do, and there are also a large number of ready-made products. But for a national-level application like Alipay, the two biggest challenges are that there are still certain user learning barriers to the use of ❶ conditional screening, and ❷ This function may be too deep hidden and difficult to find.
With an AI assistant to understand the user's natural language requirements and understand which function corresponds to, it is obviously much easier for the user to use. This matter may be more important to users than all kinds of fancy features.
In contrast, the AI assistants of various companies have a similar idea.
The same problem is thrown to the intelligent assistant "Xiaotian" of Tiantian Fund, which will also jump to the base selection module, but it needs to set the conditions from scratch, and the degree of intelligence is slightly inferior.
China Merchants Bank's AI wealth assistant "Xiao Zhao" is similar, and it is also the entrance to give corresponding functions after understanding the needs.
Of course, the AI-empowered investment base was just started last year. Even if it is a good performer, when I ask to recommend an index fund, I can't give a reference answer. As a professional fund analyst, of course, I know that such funds must exist, but it may not be included in Alipay's base selection model. From this point of view, there is still work to be done.
How valuable is the AI assistant for fund investment?
Looking at it now, Xiaohe is showing sharp corners.
I believe that with the empowerment of more professional technology and professional information, AI assistants can become a new entrance for the people to obtain services, without remembering the cumbersome functional levels, everything can be obtained with text and even voice, which is a beautiful new era for the people who do not have much time to study the functions of the fund.