The Australian Financial Review recently released an article with the headline, “A robot beat Australia’s top fundies (but it still can’t count)”. The headline, while catchy and arguably provocative, demands a closer scrutiny. The underlying story is far more nuanced than the title suggests.
Understanding the Numbers
From May 11 to August 30, 2023:
- Bard’s portfolio of 19 holdings appreciated by 8.2%.
- ChatGPT’s portfolio of 23 holdings rose 4.21%.
- The average return of the active funds chosen by Kylie Purcell was 6.3%.
- The S&P 500 rose 9.3%.
- The S&P/ASX 200 declined 0.63%.
Evaluating the Experiment
The experiment sought to determine if AI could replicate, if not outdo, the investment strategies of top active Australian fund managers. Purcell set parameters such as sustainable competitive advantages, management track records, and trading below intrinsic values. Yet, the results are inconclusive at best.
- Data Limitations: ChatGPT’s stock suggestions were based on data up to September 2021, making some of its information potentially outdated.
- AI’s Understanding: Bard, one of the AI models tested, couldn’t even form a portfolio of 23 stocks when asked multiple times. It maxed out at 19. This demonstrates that while AIs can process vast amounts of information, their fundamental understanding or execution isn’t flawless.
- Randomness: During the experiment, one large language model (Bard) beat the active fund managers, while another large language model (ChatGPT) failed to beat the fund managers. Is this result meaningful in any way over 3 months? No.
Can ChatGPT and Google Bard Pick Stocks?
While the AFR’s headline suggests a robot outperformed human fund managers, the reality is multi-faceted. The experiment does not definitively prove that AI is superior in stock picking, but it showcases both its potentials and limitations. Like any tool, AI’s efficacy in finance will depend on how it’s wielded and integrated with human intuition and expertise.
Sign Up To Our Free Newsletter