Skip to main content

Despite 540-to-1 odds, a swarm AI system correctly predicted the Kentucky Derby ‘superfecta’

Kentucky Derby prediction
Bill Brine / Flickr

If two minds are better than one, then three minds are better than two, four minds are better than three, and so on. That’s the logic behind swarm intelligence – a form of crowdsourcing inspired by nature. You can witness the power of swarms throughout the animal kingdom: birds swarm to evade predators, and ants swarm to remove debris or defend their nest. But swarms aren’t used by humans because we haven’t evolved to leverage them – until now.

A startup called Unanimous AI has developed an insect swarm-inspired system, in which individuals within a group influence each other’s decision making. In theory, a swarm makes better choices than an individual. Unanimous AI’s system, UNU, seems to depict this in practice as well.

Related Videos

Challenged by a TechRepublic journalist, Unanimous AI asked 20 knowledgable Kentucky Derby fans to use the startup’s swarm intelligence program to predict the Derby’s top four finishers. The odds of correctly predicting the top four finishers, in order –known as a superfecta– were 540 to 1. That means a $100 superfecta bet would pay out $54,000. None of Churchill Downs Racetrack’s experts put the correct horse first, but the UNU users did.

“Personally, I was speechless,” Unanimous AI CEO, Louis Rosenberg, said in a press release. “We’ve been blown away by how smart UNU has been in prior predictions, but when the horses crossed the line I almost didn’t believe it… And here’s the amazing thing – while the Swarm AI got the picks perfect, not a single individual who participated in the swarm got the picks right on their own – not one.” 

This isn’t the first time Unanimous AI’s users have swarmed together to outperform experts – since Rosenberg founded the startup in 2014, the system has correctly predicted Super Bowl and Oscar winners as well. Last year, seven non-expert UNU users correctly predicted 11 out of 15 Oscar category winners. The swarm’s 73 percent accuracy topped the New York Times’ experts by 18 percent.

Editors' Recommendations

Who does artificial intelligence think will win at the Oscars?
oscars 2016 winners and losers in every category performance by an actor a leading role leonardo dicaprio  the revenant

Awards in Tinseltown may be something of a crapshoot, but that doesn't mean that they're totally unpredictable. At least, not with the advent of artificial intelligence systems. In honor of the impending Oscars, Unanimous A.I. pooled the ever-perceptive general populace, and has come up with its predictions for the six major award categories.

So who are the big winners?

Read more
The next big thing in science is already in your pocket
A researcher looks at a protein diagram on his monitor

Supercomputers are an essential part of modern science. By crunching numbers and performing calculations that would take eons for us humans to complete by ourselves, they help us do things that would otherwise be impossible, like predicting hurricane flight paths, simulating nuclear disasters, or modeling how experimental drugs might effect human cells. But that computing power comes at a price -- literally. Supercomputer-dependent research is notoriously expensive. It's not uncommon for research institutions to pay upward of $1,000 for a single hour of supercomputer use, and sometimes more, depending on the hardware that's required.

But lately, rather than relying on big, expensive supercomputers, more and more scientists are turning to a different method for their number-crunching needs: distributed supercomputing. You've probably heard of this before. Instead of relying on a single, centralized computer to perform a given task, this crowdsourced style of computing draws computational power from a distributed network of volunteers, typically by running special software on home PCs or smartphones. Individually, these volunteer computers aren't particularly powerful, but if you string enough of them together, their collective power can easily eclipse that of any centralized supercomputer -- and often for a fraction of the cost.

Read more
Why AI will never rule the world
image depicting AI, with neurons branching out from humanoid head

Call it the Skynet hypothesis, Artificial General Intelligence, or the advent of the Singularity -- for years, AI experts and non-experts alike have fretted (and, for a small group, celebrated) the idea that artificial intelligence may one day become smarter than humans.

According to the theory, advances in AI -- specifically of the machine learning type that's able to take on new information and rewrite its code accordingly -- will eventually catch up with the wetware of the biological brain. In this interpretation of events, every AI advance from Jeopardy-winning IBM machines to the massive AI language model GPT-3 is taking humanity one step closer to an existential threat. We're literally building our soon-to-be-sentient successors.

Read more