This Predictive AI Guessed the Oscar Winners and Will Guess the U.S. President Next

[SecularMotion]

A Silicon Valley startup called Unanimous A.I. has built a methodology for gathering poll results based on public opinion, using predictive artificial intelligence to guess the decisions that people will make. The AI in question is named UNU and has already made successful guesses in the past when it comes to last year’s Academy Award winners and the Kentucky Derby winners. Next up, UNU plans to predict the outcome of the U.S. presidential election.

But how could an AI predict the outcome of an event like a horserace or an election? UNU uses “swarm intelligence,” which means that it analyzes online behavior and compiles all of that information to guess at a probable result. In the case of horse racing, UNU collated a lot of educated guesses from fans and experts and derived a prediction about which horse was likely to win. According to Engadget‘s story about UNU, only 20 people participated in the Kentucky Derby decision, but their knowledge combined still helped UNU to guess correctly.

Sounds pretty useful when it comes to placing bets on horse-racing, huh? UNU’s inventor Louis Rosenberg had the same idea, because he won $10,842 after placing a $20 bet on the Kentucky Derby based on UNU’s prediction. But that’s not the only point of creating predictive technology, although it does seem like a pretty sweet kick-back. Rosenberg explained to Engadget that he also thinks this AI could be used for more complicated problems, like diagnosing diseases:

http://www.themarysue.com/predictive-ai-election/

Who will be the next president of the US?

http://go.unu.ai/r/41384

[Bucky]

1. Gary Johnson

[jberryhill]

2. "Sounds pretty useful when it comes to placing bets on horse-racing, huh?"

No, it doesn't sound at all useful for placing bets on horse racing.

First:

"In the case of horse racing, UNU collated a lot of educated guesses from fans and experts..."

Yes, that's called "reading the tote board at the track."

The favorite - i.e. the horse with the most money bet on it - wins 1/3 of the time.

Determining which horse is likely to win the race is only half of the problem of betting on horses. You are not betting on which horse will win the race. You are betting against a market of opinions about which horse is going to win the race.

Even if there was a system that predicted the winner of the horse race with 99% accuracy, the only money to be made would be in determining when that system was most likely to be wrong 1% of the time. In the case of major stakes races, there is a lot of "dumb money" which provides opportunities, which is why the input to this system was limited to people with expertise.