Purcell asked the platform to create an equally weighted, 23-share portfolio of businesses that have a sustainable competitive advantage, strong management track record and financial inclusion, and that are trading below their intrinsic values.
ChatGPT’s stock recommendations were based on data up to September 2021, but Bard accesses up-to-date data. Purcell wanted 23 holdings, because that was the median number from 10 funds, excluding outliers.
“Bard focused more on large-cap technology stocks, but excluded cash … and I think the excellent performance of the Bard portfolio is a reflection of the market at the moment,” Purcell says.
“ChatGPT was more accurate when it came to reflecting fund managers’ strategies and a little less risk and a little more diversification in the market.”
Between 11 May to 30 August 2023:
- Bard’s portfolio of 19 holdings rose 8.2 percent.
- ChatGPT’s portfolio of 23 holdings rose 4.21 percent.
- Active funds rose an average of 6.3 percent.
- SAndThe P500 rose 9.3 percent.
- SAndThe P/ASX 200 fell 0.63 per cent.
We’re a long way from using AI to build investment portfolios, for at least one reason: Bard can’t measure. Asked several times, Bard couldn’t produce a portfolio of 23 stocks, Purcell says. Instead, the closest they could get was 19.
“We had the same problem, broadly speaking, in getting a certain level of exposure to European equities,” says Purcell.
But it goes beyond that, says Michael Kolo, former fund manager and CEO of AI consulting firm Evolve Reasoning.
Simply put, big language models like ChatGPT and Bard have information, but they can’t predict the future.
“The way we can actually build a portfolio that beats the market is to try to find things that people don’t know yet, maybe by thinking about what earnings announcements will look like,” he says.
To do that, you need to generate predictions through empirical data or qualitative storytelling.
“None of this is what big-language models do,” he says. But, he added, active managers cannot predict the future.
A more useful way to use AI is to analyze sentiment around a company and how excited or disappointed investors are likely to be about it in the future, he says. Because the way sentiment is measured and recorded, such as news articles, broker notes, and more, requires a lot of time and research.
“If you want to use these big-language models to track safety sentiment and subject sentiment (like views on whether a company is a polluter) … that’s where these kinds of models can be very powerful. ,” he says.
Financial adviser Tim McKay says investors should be careful when using an AI app like ChatGPT because – in his experience – it frequently provides incorrect or outdated information.
It is looking at investments (based on information) available two years ago. You’re driving, looking in the rearview mirror.”
Noting AI’s shortcomings, Australian Shareholders’ Association chief executive Rachel Waterhouse believes AI won’t spell the end of stock picking, but it could change the game.
Because if everyone uses the same information, they will make the same decisions. Then the quality of the prompts, the AI used and the investors’ ability to analyze the information will make a difference.
She also believes that while AI’s ability to predict investment collapses is limited, it is possible that AI itself could trigger a black swan event.
“If everyone is so dependent on AI technology to make their decisions and it’s telling us something should sell, and then everyone sells it — that’s the part I was thinking about.”