How To Invest

AI outperforms analysts: Can AI help investors beat the market?

Beating the market has long been the ultimate goal for investors who pursue active investing. As artificial intelligence (AI) continues to make significant strides across various industries, it’s now making its mark in financial analysis and stock picking. This begs the question: Can AI truly help investors outperform the market? The following article will discuss the different tools, the possibility of beating the market, and the risks to be noted.

Generative AI

The application of AI in finance comes in various forms, from generative AI (Gen AI) models to specialised stock-picking platforms. Generative AI models like ChatGPT by OpenAI, Gemini by Google and Copilot by Microsoft have garnered tremendous attention for their ability to process and generate human-like text. ChatGPT reached 1 million users in just five days, compared to Instagram’s two months. In the same two-month span, ChatGPT soared to 100 million users. Although Gen AI models like ChatGPT show promise, their application in stock picking reveals some significant limitations.

When the models were asked to provide ‘top 10 stocks ideas with strongest fundamentals and largest margin of safety’, these models responded differently. ChatGPT offered a list with detailed information, Copilot provided a similar list but with less detail, while Gemini refrained from giving recommendations, instead suggesting ways to develop investment knowledge. When asked for specific financial metrics like CAGR growth or intrinsic value, AI hallucinations come into play, and a fact check on the numbers is often needed.

This variance in responses highlights a key challenge with generative AI: these models are trained on historical data, often lack access to real-time market information, adhere to platform guidelines that limit certain disclosures, and are prone to generating hallucinations. They also may not provide current or actionable investment advice, making them less reliable for making specific investment decisions. 

Specialised AI investment platforms

In contrast to standard generative AI, specialised platforms like Danelfin and Boosted.ai are explicitly designed for investment analysis stock picking. These platforms continuously update their data and algorithms to provide more current and relevant insights. They also integrate real-time financial data to minimize the risk of AI hallucinations.

For example, Boosted.ai will continually read news articles, company filings, earnings reports, and other sources of information for their self-created Artificial Intelligence Index. At the moment, the platform caters to advisors and institutional investors, enabling them to leverage machine learning in their portfolios to optimize investment strategies.

On the other hand, Danelfin caters to retail investors, and claims its AI can outperform the S&P 500 in the short term. The platform uses a combination of technical, fundamental, and sentiment indicators to generate ‘AI scores’ for stocks and ETFs.

Recent studies on AI stock performance

A recent study from the University of Chicago compared the performance of AI models to human analysts. Researchers used ChatGPT 4.0 Turbo to analyse standardised financial statements of over 15,000 companies from 1968 to 2021.

A Chain-of-Thought (CoT) prompt was developed to effectively ‘teach’ the AI model to mimic the reasoning process of a financial analyst. The CoT prompt replicates this analytical process through a structured set of instructions, ultimately guiding the model to determine whether next year’s earnings will increase or decrease compared to the current year.

The results were intriguing: A non-CoT prompt in GPT-based forecasts delivered a performance of 52%, which fell short of analyst benchmarks, ranging from 53% to 57% accuracy. However, when the CoT prompt was employed to replicate human reasoning, the AI achieved an accuracy of 60%, significantly surpassing the analysts’ performance.

Accuracy
Random Walk 49.11%
Analyst 1m52.71%
Analyst 3m55.95%
Analyst 6m56.68%
GPT (without CoT) 52.33%
GPT (with CoT)60.35%
Prediction performance of the random walk model, analysts’ forecast issued one month after previous earnings
release (Analyst 1m), three months after previous earnings release (Analyst 3m), and six months after previous earnings release
(Analyst 6m). GPT (wihtout CoT) denotes GPT’s predictions without any chain-of-thought prompts. Accuracy is the percentage of correct predictions out of total predictions. Source: University of Chicago

However, it’s not all bad news for human analysts. According to the same research, humans still hold a crucial advantage during economic shocks, such as the 2008 financial crisis or the COVID-19 pandemic.  This is because humans can better interpret and react to unprecedented events and external factors that AI models may not be designed to account for.

Potential risks and considerations

While the potential of AI in investing is exciting, it’s crucial to consider its limitations. One significant drawback is AI’s inability to factor in real-time events effectively. Most AI models are trained on historical data and may struggle to adapt to breaking news or unexpected market shifts. Besides, AI lacks the intuition and experience that seasoned investors bring. While AI excels at pattern recognition, it can’t replicate the nuanced understanding of market sentiment or the broader economic context that humans possess (for now).  

Therefore, it is essential to view AI as a tool rather than a magic solution. The above study also found that combining GPT-4’s predictions with human forecasts actually enhances overall accuracy. Hence, investors should use AI-generated insights to complement their own analysis and decision-making process rather than replace it entirely.

The fifth perspective

Looking ahead, AI holds significant potential to democratize financial analysis, though it remains in its early stages within the investment world. As technology advances, a hybrid approach is likely to emerge, with AI complementing rather than replacing human expertise. AI could take on time-consuming tasks like data analysis or summarizing lengthy transcripts, while tools such as an “AI Score” can help validate our insights. This allows investors to focus more on interpretation and strategic decision-making. Notably, established financial data providers like Bloomberg Terminal, Capital IQ, and FactSet are already integrating AI capabilities into their platforms.

It is essential to view AI as one of many tools rather than a standalone solution. Therefore, understanding both its potential and limitations is key to effectively leveraging it to enhance our analysis and decision-making. As AI evolves, it continues to be a fascinating new player in the age-old pursuit of beating the market.

Darren Yeo

Darren Yeo is an investment analyst at The Fifth Person, where he provides insightful analysis to help readers make more informed investment decisions. Before joining The Fifth Person, Darren gained two years of experience working at a bank. With a keen interest in finance, he is dedicated to continuous learning in the field of investing.

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button