Can AI predict the Melbourne Cup winner? I decided to find out

ChatGPT can create meal plans, draft an essay and summarise lengthy reports. It can simulate therapy sessions, solve complex maths problems and give you relationship advice.

(Though if you need ChatGPT for relationship advice, it’s perhaps a relationship beyond saving).

Generative AI might be a relatively new technology, but it’s already embedded across our workplaces, our smartphones and our classrooms. It’s relied upon for being smarter and quicker than humans at many everyday tasks.

Melbourne Cup winner Knight’s Choice (right) wasn’t AI’s choice in the tipping stakes.Credit: AP

And there are heady promises of what AI can do in the future, too. The jobs it will create, the menial jobs it might take away and the billions in economic growth it will create through efficiencies and automation. As technology editor, I get no shortage of pitches in my inbox about how we’ll soon be relying on generative AI for everything, and that it will inevitably upend our economy and disrupt every sector and industry.

But can it tip a Melbourne Cup winner? Well, I’d sure hoped so.

Though I’m a self-confessed sport diehard, I’m not exactly the world’s best tipster.

People talk about betting with your head, not your heart. But I’m firmly a “heart” guy. I always tip my beloved Sydney Swans, and typically back a horse based on name alone: Fancy Man would have been my firm bet for this year’s Cup. I haven’t had much luck over the years.

Any technological system built on databases and information, however unreliable it still might be, could only be better than my non-system.

To put that thesis to the test, I consulted three different generative AI systems: ChatGPT, Google Gemini and the Australian-made PunterGPT, and put a couple of bucks on each of their predicted Cup winners.

PunterGPT’s tip

PunterGPT is the brainchild of Michael Kollo, chief executive of generative AI company Evolved Reasoning, and John Conomos, the head of global quantitative research at Macquarie Group.

It has form, last year predicting eventual winner Without a Fight to finish in the top three.

PunterGPT’s tips for the Melbourne Cup.

PunterGPT’s point of difference is that it takes in words, not numbers, according to Kollo. It’s been fed a cross-section of different articles written by different professional punters, with the view that in among all of that expertise, there is a truth. Its whole shtick is about picking up sentiments and perceptions of a particular horse, and comparing that to the market odds.

It’s no surprise that, given how niche and specialised PunterGPT’s expertise is, it’s the most bullish of the lot. It tipped relative outsider Onesmoothoperator at $8.50.

An AI-generated image of the Melbourne Cup from PunterGPT.

“His win in the Geelong Cup was eye-catching,” the artificial intelligence says.

“Coming from the back, running wide and still showing a powerful finish tells us he’s got staying power and a turn of foot, too.

“With his light weight and a top jockey in Craig Williams, he’s got the tools to make a big play at Flemington. Plus, he’s not just consistent but has been keeping up in these staying races, which suggests he’ll stay the 3200m. Definitely a top chance in my book and good value.”

I heed PunterGPT’s advice and put a couple of dollars on Onesmoothoperator.

ChatGPT and Google’s Gemini are much more generalised in how they go about it, so are more circumspect.

ChatGPT’s tip

“If I had to pick just one horse for the 2024 Melbourne Cup, Buckaroo is a strong choice,” ChatGPT says.

“He’s currently a co-favourite and comes off a solid performance, finishing as runner-up in the Caulfield Cup, which showcases his ability in high-stakes races. His consistency and recent form make him one of the most reliable options, especially if you’re looking for a straightforward bet rather than taking a chance on a dark horse.”

I’ve already backed somewhat of a dark horse in Onesmoothoperator, so I put a couple of buckaroos on ChatGPT’s pick.

Google Gemini’s tip

Google’s generative AI chatbot Gemini is downright cowardly in comparison.

“As an Al, I cannot provide personal opinions or beliefs,” it tells me. “However, based on the pre-race analysis and expert opinions, Vauban seems to be a strong contender for the 2024 Melbourne Cup.”

A tip is a tip. I round out my betting with a couple of bucks on race favourite Vauban, even though an outright favourite hasn’t won the Melbourne Cup in more than a decade.

The result

AI might be decent at generating a meal plan or solving a complex maths problem. It might even be OK at giving relationship advice (though really, please see a couples’ counsellor before turning to ChatGPT).

As far as horse racing goes, though?

Generative AI was a resounding failure, Knight’s Choice taking out the Melbourne Cup, followed by Warp Speed and Okita Soushi. None of the horses tipped by AI finished in the top five, and my betting yielded exactly zero dollars in winnings.

For now, I’m going to stick to my own non-system. At least it has some heart.

Get news and reviews on technology, gadgets and gaming in our Technology newsletter every Friday. Sign up here.