Want to save yourself a few hours of elevated heart rate this Sunday? Allow us to unceremoniously spoil the result of the much anticipated Kansas City Chiefs vs. San Francisco 49ers game. In what is genuinely one of the biggest toss-ups in recent Super Bowl history, you can take it to the bank that Kansas City has it in the bag. Well, at least sort-of in the bag.
Let’s put it this way: If the two teams were to play 10 games against one another, the Chiefs would probably win six of them. And as good as the 49ers’ defense is, the Chiefs’ offense is just that much better. The result will be an exciting, high-scoring game with total points exceeding 55, that sees the Chiefs eke out a victory after an early lead by the 49ers. The only way the 49ers realistically win this thing is if they can build up enough of a lead in the first half, then eat up the clock with the running game. Or something like that.
The fact that we have Super Bowl predictions for you is not, of course, unique. Ask any taxi driver, walk into any barber shop, visit any website (I mean, don’t actually; just take my word for it) and you’ll struggle to not hear some strongly worded opinion about who has this year’s Super Bowl sewn-up. Opinions, as they say, are like assholes. Everyone has one and most of them stink.
But Unanimous AI‘s opinions are different than most. The artificial intelligence company, founded by former Stanford computer scientist Dr. Louis Rosenberg uses swarm intelligence algorithms — based on the responses of many, many people — to create its predictions. But while that theoretically means a whole lot more assholes, the results surprisingly smell far sweeter.
A different type of artificial intelligence
“Unanimous A.I. is a very different type of artificial intelligence company as we don’t replace humans with algorithms, but instead use A.I. to amplify the knowledge, wisdom and insights of human groups,” Rosenberg told Digital Trends. “In layman’s terms, we build super-intelligent ‘hive minds’ by connecting groups of people over the internet, enabling them to think together as real-time systems moderated by A.I. algorithms. In technical terms, the technology is called artificial swarm intelligence. That’s because it’s modeled on how swarms in nature amplify the intelligence of groups.”
As Rosenberg notes, in the natural world many species amplify their intelligence by forming real-time systems. We don’t call these things real-time systems, of course. If you’re a regular person, you tend to refer to them as flocks, schools or swarms: large collectives of animals joining forces to create some kind of mass intelligence that is both the sum of, and weirdly independent from, their individual selves.
These collectives can often look random, but, according to Rosenberg, they’re not. “They make significantly more accurate decisions when thinking together in systems than they could on their own,” he said. “This was the motivation to found Unanimous A.I. back in 2014. After all, if birds and bees and fish can get so much smarter together, it should work for humans too if we can create the right technology to connect people together.”
Rosenberg and colleagues got together and built a software platform called Swarm.ai. This platform invited groups of humans to make decisions, predictions, and forecasts, and tested to see whether or not this had the effect of amplifying human predictive abilities. This system works by asking knowledgeable people to log on and then attempt to move their cursor in the direction of the answer to a question (like who is going to win a particular sports game.) The A.I. algorithms infer things based on these answers, not just in terms of the answer, but also the conviction with which they’re given.
Predicting marketing campaigns to sport
This approach doesn’t just have to be about sport. In a 2018 study, Unanimous A.I. partnered with researchers from Stanford University Medical School to test whether groups of doctors, when thinking together as a “hive mind,” made more accurate diagnoses than traditional methods. The improvement wasn’t marginal, either. It was by a massive 33% in terms of reducing diagnostic errors.
Another demonstration, carried out in conjunction with MIT researchers, asked groups of financial traders to predict the change in price of financial assets such as gold, oil, and the S&P 500. Again, forecasting accuracy was improved by a similar number — in this instance, 36% compared to traditional forecasting.
Other tests of Unanimous’ swarm technology have looked at everything from Oscar predictions to helping Fortune 500 companies predict the impact of product price changes. The sports business is where it’s really gained interest, however.
“It should be noted that when we first got into sports forecasting, we did it because it was simply a very good testbed for confirming whether or not our technology was enabling us to amplify the intelligence of human groups at a significant level,” Rosenberg said. “After rigorously showing strong results when predicting basketball, soccer, hockey, and football in these studies, Unanimous A.I. got flooded with interest from sports fans who wanted access to our predictions on a regular basis.”
Out of the goodness of their heart, or potentially sensing a lucrative business opportunity. Unanimous A.I. last year launched a spinoff company called Sportspicker A.I. This business uses swarm intelligence to generate daily predictions about a variety of professional sports, including football, soccer, hockey, basketball, and baseball.
To bet or not to bet?
“The most recent data we have is for the NFL playoffs, which comprised a set of 10 very competitive games,” Rosenberg continued. “As published to our subscribers, Sportspicker A.I. correctly predicted the outcome, against the spread, of 7 of those 10 games. Had somebody started with a $100 bankroll and placed wagers on each of those 10 games as forecast by Sportspicker A.I., they would have ended up with $170.”
That, he notes with a justified degree of pride, is approximately a 70 percent return-on-investment. (Although, unless my math is very much off, the subscription cost for the service will make a bit of a dent in those winnings.)
As with any kind of prediction, nothing is guaranteed. Swarm A.I. may lay claim to picks that aren’t just based on the “gut reaction” of a few paid handicappers, but are instead “data driven, scientific, and objective.” But anywhere the messiness of human unpredictability can sway results, there’s no guarantee that data, science, and the claim to objectivity will be enough to provide the correct result every time. It is perhaps for this reason that, unlike previous years, Unanimous A.I. has not predicted an exact score this year.
On balance, maybe you’re better off tuning in to watch the Super Bowl live, after all. Even if you do come away wishing you’d bet the house on the Kansas City Chiefs.
- OpenAI’s GPT-3 algorithm is here, and it’s freakishly good at sounding human
- The history of the Madden Curse
- Facebook study shows how social media may influence our vacation behavior
- A literal minority report: Examining the algorithmic bias of predictive policing
- Microsoft reportedly turns to A.I. to optimize MSN news, replacing human workers