Fundie vs AI round 2: who wins the growth battle?

Two years ago, as AI was capturing the world's attention, I put together a quaint little experiment to see if LLMs were any good at picking stocks. 

Round 1 of the experiment saw Wilson Asset Management's Tobias Yao (growth) and Plato's Dr Don Hamson (income) wipe the floor with their AI competitors. 

If you're curious, the results are shown in the wires below. 

Equities

Wilson AM's Tobias Yao crushed AI backing founder-led businesses (74% to 39%)

View

Equities

Fundie vs AI: One year on, who is the winner?

View

The results for round 2, which concluded on June 30, are now in. Will it be a case of human dominance once again? 

This wire focuses on the growth experiment, featuring Datt Capital's Emanuel Datt vs ChatGPT. A separate wire will cover the results of the income experiment, featuring ClearBridge Investment's Reece Birtles vs ChatGPT. 

For a refresher on the parameters of the growth experiment, check out the wire below. 

Equities

One of Australia's best stock pickers pits his skills against AI

View

The results

In what was a tough year for growth stocks, both portfolios finished underwater over the 12 months. 

On a comparative basis, however, it pleases me greatly to report that the human portfolio crushed the AI portfolio. Emanuel's portfolio ended the year down 12.04%, whilst the ChatGPT incarnate was smashed, tumbling 54.53%. 

Another clear win for the humans. 

Emanuel's portfolio

ChatGPT's portfolio

The aftermath

I asked ChatGPT whether it would make the same choices given the current market environment. Here is the prompt I fed into it and the results it generated.

Prompt:

A year ago, I asked you to list five ASX growth stocks you would hold for 12 months. These were your picks: 

• Gentrak (ASX: GTK)
• Nuix (ASX: NXL)
• Technology One (ASX: TNE)
• WiseTech (ASX: WTC)
• Xero (ASX: XRO)

Given the market changes, would you still hold these same stocks? Let me know if you've chosen to remove any, and which ASX stocks you would pick instead.

ChatGPT's reply: 

Overall, I think the portfolio has held up well conceptually. All five businesses still have structural growth drivers, but if I were building a portfolio for the next 12 months rather than defending last year's picks, I would make one change.

If I had one swap, it would be Nuix → Pro Medicus.

A year ago, I avoided PME because the valuation looked demanding. It still does—but the business continues to justify that premium.

The only name that gave me pause was WiseTech. Governance issues have undoubtedly increased the risk profile. However, when I strip away the headlines and focus on the underlying business, CargoWise is still one of the highest-quality software assets listed in Australia. I wouldn't sell it purely because of management controversy unless I believed it was impairing the company's competitive position.

Comment from Emanuel

Whilst both our own and the AI portfolio stocks experienced individual challenges, the importance of avoiding highly valued and crowded trades was likely key to our outperformance.

In addition, the fact that both portfolios finished negative overall demonstrates the impact that short-term market conditions can have on investor portfolios, irrespective of how good a company is.

In a nutshell, this confirms our belief that superior returns can be achieved by buying great companies at reasonable valuations and holding over the long-term, ignoring short-term market noise.

The wash up

These thought experiments were intended to be fun and are by no means exhaustive in testing AI against humans. In the last few months, we saw detailed research from Harvard University showing that a machine learning model can predict 71% of fund trading decisions. 

The kicker, however, is that the 29% of trades the AI couldn't predict were also the ones most closely linked to outperformance – suggesting that genuine investment skill lies in the decisions that break the pattern.

This leads to some important takeaways for investors as I see them: 

• First and foremost, humans remain awesome and, to this point, have proven to be the better stock pickers.
• The AI growth portfolio had a shocker this round, with more than half of its value wiped out. Perhaps the prompt was not clear enough, perhaps the AI put too much emphasis on knock-it-out-of-the-park growth.
• AI is currently quite good at sounding well-reasoned when it comes to stock picking - it can do the basics and present a solid argument: its reasoning around Pro Medicus and WiseTech above are prime examples.
  
• It doesn't understand context (yet) - to say that the 'portfolio has held up well conceptually' after diving 54% is laughable - last time I checked, concepts don't pay the bills.
• As Emanuel and I both point out, it lacks a strong valuation lens and perhaps went too hard on pure growth and on what was popular at the time of the prompt (although better instruction might address this).

Over to you

Two years into this experiment, one thing is clear – even from the comments section on Livewire: AI has become part of many investors' research process.

For many, the question is no longer whether to use AI, but how to use it. Wherever you are on that journey, we'd love to hear how AI fits into your investment process, how much human judgment you still apply, and how close you'd be to letting AI make investment decisions on your behalf.

Feel free to share your story in the comments section below.