Can AI bots like ChatGPT make us better investors?
Trading platform Tiger Brokers has released an investment-focused AI chatbot, so we gave it a spin.
A couple of months ago, Finder ran an experiment that showed an AI-generated stock portfolio outperforming Australia's 10 most popular actively managed funds over a period of 3 months.
While the test doesn't prove we should all be using AI instead of financial advisors, it does raise some interesting questions around what might be possible.
It also highlighted some of the shortcomings of AI bots, such as outdated data - ChatGPT can only access data to January 2022 - and the inability to accurately count, among other things.
Now Singapore-based online trading platform Tiger Brokers has launched its own AI chatbot TigerGPT that might solve some of these problems.
Using ChatGPT technology but specifically designed to problem solve investment queries, it draws on subscription-only data, live prices and news. It also pulls from contributor analysis, making some of its data entirely unique to Tiger customers.
Ahead of its launch, I was given access to TigerGPT, so I thought I'd give it a spin to find out how it fared against ChatGPT and if it really could help me to be a better investor.
How to sign up to TigerGPT
If you want to try out TigerGPT for yourself, you'll need to sign up to the Tiger Trade trading app - you can do that from our Tiger Brokers review by clicking on the 'Go to Site' button. Once you've signed up, go to the home screen and click on the blue chatbot symbol on the bottom right of the screen.
The results
Here's a quick summary of what I found when comparing ChatGPT with TigerGPT.
To get more accurate responses, I also tested ChatGPT's web plugin which allows it to collect answers in web search results.
- Live prices and price performance: TigerGPT wins by a large margin
- P/E ratio and fundamental data: TigerGPT wins
- Technical analysis: ChatGPT wins - but only with a web plugin switched on
- Upcoming dividend information: ChatGPT + web plugin wins
- Broker recommendations: Only ChatGPT + web plugin could provide individual broker recommendations (for example from Goldman Sachs or Morgans) and price targets for stocks. TigerGPT was only able to provide recommendation averages across multiple brokers.
- Final buy recommendation: Both could reach a conclusion on whether a stock is a good buy, however ChatGPT + plugin provided more supporting evidence.
Stock prices and fundamentals
With direct access to live data, TigerGPT has a clear advantage over ChatGPT and similar AI bots like Bard when it comes to gathering market data information.
In requesting the live share price of ASX:CBA, only TigerGPT was able to give a correct response. Tiger was also the only model able to give accurate asset price changes over specific time frames as well as fundamental data such as P/E ratios and debt-to-equity.
ChatGPT on the other hand was unable to give an answer even with the web plugin switched on:
I'm sorry, but I cannot provide real-time stock prices. To get the live share price of ASX:CBA (Commonwealth Bank of Australia) or any other stock, you'd need to check a financial news website, use a stock market app, or visit the official website of the Australian Securities Exchange (ASX).
It's worth noting that live stock data is a relatively closely guarded commodity. Even Google's Bard, which has access to Google Finance data, would only give CBA's share price from the previous day, rather than the current price.
Stock recommendations
Of course, the real question is whether AI can actually use all this to data to help us make profitable investment decisions.
I asked both ChatGPT and TigerGPT "Is Tesla stock a buy?" adding for nuance "for an investor with a high risk tolerance and a long time horizon."
ChatGPT with the web plugin switched on actually gave a pretty detailed analysis of around 600 words.
Firstly, Tesla is often characterized as a high-risk, high-reward stock. The company has been successful in capturing significant market share in the electric vehicle (EV) sector, reaching 14.55% of all worldwide sales in the first half of 2021. However, there are several risks and challenges that it faces over the next 5-10 years.
It was able to highlight key issues facing the carmaker such as the high price of Tesla vehicles, inability to meet demand for battery supplies, growing competition in the EV space and scandals surrounding CEO Elon Musk.
It also provided a detailed analysis of TSLA's price volatility using technical and fundamental analysis to reach a conclusion:
Given these considerations, a high-risk tolerant investor could see Tesla as a potentially rewarding long-term investment, given the company's leading position in the EV market and the industry's growth prospects. However, due to the company's risks and stock volatility, such an investor should be prepared for potential periods of price fluctuations and performance uncertainty.
TigerGPT presented less analysis (around 150 words) and relied more on broker recommendations and average price targets to reach its conclusion. It did not include technical analysis indicators or expand on Tesla's business prospects.
There are mixed opinions about Tesla stock from different analysts and articles. The mean, low, and high target estimates range from 24.333 USD to 350.0 USD. 6 analysts recommend buying the stock, 11 analysts suggest holding it, while 7 analysts recommend selling the stock.
It ultimately reached a similar conclusion as ChatGPT:
Tesla stock is potentially a good buy for long term investors with a high risk tolerance who believe in the company's growth prospects, but it also comes with significant uncertainties.
Technical and fundamental analysis
While TigerGPT has access to more up-to-date data, it falls short on putting the pieces together.
Here's TigerGPT's response to the prompt "Provide a technical analysis of Tesla stock":
Based on the information provided, Tesla's stock price is 265.28 USD per share. The P/E ratio as of the same date is 75.175, and its market cap is 840.81B USD. The EPS is 3.53. However, without access to additional technical analysis data, it's difficult to provide a thorough technical analysis of Tesla. Therefore, I cannot provide a comprehensive technical analysis of Tesla without more relevant information.
It also had trouble with financial info that isn't readily available in data feeds. For instance, when asked for the dividend payment date of various stocks, it gave a past date.
When asked for stock ratings from specific brokers such as Goldman Sachs or Morgans it was unable to give an answer.
While ChatGPT faced similar issues, its web plugin was able to provide it with more useful information on a range of technical indicators including moving averages, bollinger bands and relative strength index.
Price and Moving Averages: The current share price of Tesla is $269.06. The stock's 8-day and 20-day simple moving averages (SMA) are higher than the current price, suggesting a sell signal. However, the 50-day and 200-day SMAs are lower than the current price, indicating a buy signal. A similar trend is observed with the 8-day, 20-day, 50-day, and 200-day exponential moving averages (EMA).
ChatGPT's summary on Morgan Stanley's price target for CBA:
The latest price target set by Morgan Stanley for CBA is $79. This is a reduction from the previous target of $91.
The broker's new price target implies a possible decline of more than 10% over the next year. This reduction in price target is due to Morgan Stanley's pessimistic outlook on the banking sector, with the belief that a quick and aggressive tightening cycle will lead to a weaker housing and mortgage market and a higher probability of recession.
The takeaway
Does all this mean AI Chatbots can help us to be better investors? Well, yes and no.
Both ChatGPT and Bard will be the first to tell you that they can't make predictions about the future. In fact they provide numerous risk warnings if you try to get them to do it anyway.
However it's clear AI is capable of collecting and analysing vast amounts of stock data to reach a reasonable, if somewhat vague, conclusion. And these chatbots can do it at a speed and scale impossible for humans to match.
But the test also shows the shortcomings of AI technology. Neither TigerGPT or ChatGPT could provide unique insight without accessing and citing various news articles and existing analysis.
For instance, ChatGPT was unable to respond to most of the questions without the web browser plugin switched on, suggesting many of its more nuanced responses simply reflect existing analyses by us human folk.
It's also difficult to trust the information provided. On a few occasions both chatbots confidently presented incorrect information without warning.
What they do offer is an easy way to collect useful information to help you to form your own investment conclusions. Research that might previously have taken hours can now be gathered in minutes with a few simple prompts. That in itself is a pretty powerful thing.
Important: This test was carried out between July and August 2023.
Finder’s Insights Column examines issues affecting the Australian consumer. It appears weekly on finder.com.au.