AI "stock speculation", independently managed funds, the era of real robo-advisors is coming? Following the use of ChatGPT for stock selection, a domestic fund company recently announced that an artificial intelligence-based robot (tentatively named "Cybertron") will independently manage a fund.
In fact, in the context of the AI boom sweeping the world, the topic of artificial intelligence (AI) participating in investment and serving as an investment adviser has attracted enough attention. Gao Zhiwei, chief analyst of financial engineering at Guojin Securities, told reporters that investors have used artificial intelligence models such as deep learning and reinforcement learning for quantitative investment, and artificial intelligence may not be able to achieve completely accurate prediction of the stock market, but it is possible to achieve better prediction results on broader issues such as finding a type of stock with a greater probability of rising and predicting which industries may usher in a higher prosperity.
AI "stock trading", independently managed funds
Artificial intelligence marches into the domestic investment market! On the afternoon of June 1, the private equity fund "Stop at Extreme Investment" announced that it intends to arrange four researchers and an artificial intelligence-based robot (tentatively named "Cybertron") to independently manage five different private equity funds.
It is reported that the fund managed by "Cybertan" is so far as Zhishan No. 1, which is also the first AI fund manager in the domestic market. "We have fully communicated with all the above fund investors and have been affirmed." The independent management of funds by AI robots will be our first attempt to integrate subjectivity and AI, and invest in the global market in a quantitative way, laying a good foundation for the AI era. ”
Before AI independently managed funds, some platforms had already allowed them to "speculate in stocks". A foreign trading platform called Autopilot recently carried out an experiment to try to verify the reliability of using ChatGPT to analyze the stock market and guide investment. By the end of May, 25,000 investors had been attracted and staked more than $10 million on ChatGPT's selected portfolios in order to achieve an annual return on investment of more than 500%, as claimed in a previous research report.
However, based on feedback from the past two weeks, the investment trend does not seem to be very ideal: the AI-selected stock portfolio rose by about 2%, which is basically the same as the broader market, and the bottom five stocks in the portfolio fell more than the top five in percentage terms.
Why is there such a difference from an expected return of more than five times to a realistic result of around 2%? Gao's analysis points to two reasons: On the one hand, the experiment is based on a paper by a scholar at the University of Florida, in which the portfolio that reached 500% yield did not take into account the impact of transaction costs. In fact, the paper also backtested different transaction costs, and found that the level of income decreased significantly when transaction costs increased.
On the other hand, the portfolio with a 500% return in the paper was obtained by buying the highest-scoring stocks and selling the lowest-scoring stocks at the same time, while the portfolio in the Autopilot experiment ordered ChatGPT to analyze more than 10,000 news and get the top 100 stocks with the highest scores, combined with the company's financial report data to get a comprehensive score, and finally bought the top 20 stocks. In the case of losing the short part of the gain, the actual investment performance will naturally be quite different from the backtest results in the paper.
"The quantitative strategy pursues a long-term outperformance of the market benchmark, and the actual results of only two weeks are not enough to reach a more effective conclusion, and the actual investment effect of the strategy needs to be further observed." Gao Zhiwei believes that with the gradual development of large language models such as ChatGPT, it can provide certain auxiliary decision-making roles for the field of investment research. Stock market forecasting itself is an open question, artificial intelligence may not be able to achieve completely accurate forecasting, but in a broader level such as finding a class of stocks with a greater probability of rising, predicting which industries are likely to usher in a higher prosperity, artificial intelligence is completely possible, better prediction effect is completely possible.
Is the era of robo-advisors coming?
For investors, quantitative investing is no longer a strange concept. This investment method of using computer programs and digital models, supplemented by a large amount of data, mining stock selection factors to trade has been developed for many years, and the emergence of large language models such as ChatGPT has undoubtedly injected new blood into it. "There are already investors who use AI models such as deep learning and reinforcement learning for quantitative investment, and many models have achieved good prediction results." Gao Zhiwei pointed out in the interview.
Gao said that in many quantitative investment fields, models such as ChatGPT can be reshaped. The incremental information provided by these large-language models with high independence can often complement and enhance the original quantitative investment strategy or model. With the emergence of such models, in order to further improve their capabilities in the field of quantitative investment, some quantitative investors may consider deploying and fine-tuning their own large-language models, and such models have huge requirements for computing power, which may cause fierce competition among investors in computing power investment.
From the perspective of the investment market as a whole, Gao believes that if models such as ChatGPT are widely used in future investment activities, it may be beneficial to investors and the stock market. "On the one hand, models such as ChatGPT can help investors collect and process a large amount of financial data, identify potential trading opportunities or risks, and then attract incremental funds to enter the market, improve the liquidity and effectiveness of the entire market, and the improvement of market effectiveness will also reduce excessive volatility of the stock market; On the other hand, funds that invest using models such as ChatGPT have the potential to improve their return levels and enable investors to obtain better investment returns."
"In quantitative investing, ChatGPT can complete some simple quantitative strategies, such as building an average regression model and outputting a moving average strategy. Theoretically, it is even possible to use the Scikit-learn database (a free software machine learning library for the Python programming language) to build a predictive model for future interest rates, and evaluate them using MSE (mean squared error). Xu Liang, chief engineer of the OneConnect Artificial Intelligence Research Institute, pointed out to reporters.
Xu Liang said that robo-advisory is a professional investment consulting service based on artificial intelligence, and after the application of AI robots such as ChatGPT, it can eliminate human subjective factors, provide more objective advice, and continue to evolve and innovate with changes in the market and environment. In the future, ChatGPT can be applied to natural language processing (NLP) and chatbots, predictive analytics, reinforcement learning, blockchain, computer vision and other emerging hot fields, and will have a huge impact on the development of the financial industry.
However, Lin Jianming, founder of Samoyed Cloud Technology Group, told reporters that financial institutions have huge imagination in using ChatGPT technology to gradually empower scenarios. Intelligent marketing, intelligent customer service, risk identification, code programming and other aspects are all good application directions of ChatGPT. However, it is not a good choice to use ChatGPT for quantitative investment and robo-advisory. Because ChatGPT cannot replace the interpersonal relationships and investment experience of an investment advisor, nor can it replace an investment adviser to provide personalized investment advice to the client based on their unique circumstances.
This article is from the International Finance News