Swarm AI Correctly Predicts Kentucky Derby Superfecta, Turns $20 Into $11,000 (yahoo.com)

New "Swarm Intelligence" platform UNU from Unanimous A.I. made a bet on the Kentucky Derby this weekend and won big. The bet is called the Superfecta and it paid 540 to 1 odds. "Swarm Intelligence" allows groups to amplify their collective IQ beyond the capacity of individuals, something that the human species hasn't been able to do because of evolutionary restraints. Silicon Valley startup Unanimous A.I. set out to answer one question: Can humans swarm, and if so can we amplify our intelligence beyond the ability of individuals? Spoiler: yes we can. According to Yahoo, "Unanimous spent the last two years building a swarm intelligence platform called UNU that enables groups to get together as online swarms -- combing their thoughts, opinions, and intuitions in real-time to answer questions, make predictions, reach decisions, and even play games as a unified collective intelligence." Already, UNU has predicted the Oscars better than experts, and predicted the NCAA college bowl games with 70% accuracy. As for the Kentucky Derby, Hope Reese, reporter for Tech Republic and the Atlantic, challenged Unanimous A.I. to use UNU to predict the winners. The group used UNU to answer questions as a unified Swarm Intelligence, narrowing the field of 20 horses down to four winners. Then it was asked to order the winners into Win, Place, Show, and Fourth. Swarm Intelligence convened again a week later after the Derby announced the post positions of the horses -- one of the four picks was replaced by an alternate.

Comments

Reading between the lines

[mwvdlee]

How many other bets did they place on the Kentucky Derby?

Re:Reading between the lines

[Cortexia]

That's the amazing thing... They placed ONE BET and it was done in response to a challenge by a reporter at TECH REPUBLIC, who published their pick in advance - ONE PICK. This is the article that Tech Republic published 2 days before the race: http://www.techrepublic.com/ar...

Re:Reading between the lines

[michelcolman]

Great, now all you need to do is run your algorithm in reverse!

Re:Reading between the lines

[TechyImmigrant]

>Nyquist had 2-1 odds of winning.
That's why Ts must be greater than or equal to twice the bandwidth to avoid aliasing.

Re:Reading between the lines

[bws111]

There are no 'odds makers' involved. This is parimutuel betting. The odds are determined solely by how many people are betting on a given outcome versus all the other outcomes. In other words, the 'odds' are determined by 'swarm intelligence', and have been for centuries. This is not something new.

If you want to use 'swarm intellegence' to place a bet, all you have to do is look at the current odds. Whatever has the lowest odds the 'swarm' has determined is the likely winner. I can't figure out what this great AI has supposedly done. Use 'swarm intelligence' to determine a likely outcome? That is called betting the favorites, and requires zero intelligence.

Re:Reading between the lines

[dragon-file]

It isn't amazing at all. They made a bet, and as luck would have it, it won. It was definitely a more informed bet than the average bettor would make, but I don't care how good this thing is, it will never make a living betting ponies.

Actually you might be surprised how amazing it is. I remember a statistician in high school telling me a story about those stupid guess how many beans are this jar deals. Apparently, if you gather as many answers as possible and find the average, it will be accurate within 2-5 plus or minus. Obviously the larger your sample the more accurate your results. So while I am not surprised that humans can be more intelligent in large groups, I still find it amazing.

Re:Reading between the lines

[jandrese]

Isn't it pretty rare for the oddsmakers to get it so right? If this "Swarm AI" can repeat its performance on the Preakness and Belmont then I'll start to pay attention. Thus far we don't have enough data to separate and entirely new paradigm of cognition with dumb luck.

Even a broken clock is right twice a day

[misnohmer]

One win proves absolutely nothing. Show a consistent string of wins that is otherwise statistically improbable, then actually put money into the bets (rather than theoretical betting), then you have some proof. Win only 10% of the 540 to 1 payouts and you have a money machine.

Re:Even a broken clock is right twice a day

[Troyusrex]

If this platform could reliably make bets such as this then the last thing they'd be doing is telling everyone about it. They'd use the platform to make billions for themselves betting. Once they were set for life maybe they'd share the technology.

More Collective Intelligence than Artifical

[SafPlusPlus]

Judging from skimming the links it looks to me like it's more of a social platform for collections of people to come to a conclusion. So instead of what is generally considered A.I. this would be more aptly called C.I. (as in, Collective Intelligence).

Re:More Collective Intelligence than Artifical

[Coisiche]

It looks like a newfangled name for the wisdom of crowds.

*cough* no.

[nadaou]

It always amazes me how profoundly bad very smart people can be at basic statistics. Granted it's usually not intuitive but it can be learned.

odds are collective intelligence already

[sittingnut]

bookmakers create and change odds based on the bets being placed, so that they themselves will not take any loses. that is odds are themselves the predictions of all who have paced bets. iow it is a form of collective intelligence.

whole concept of predictive markets is based on harnessing the collective intelligence/information of those who are willing to take risk on a event outcome ,to predict probability of various out comes of events https://en.wikipedia.org/wiki/...

how is this so called "swarm intelligence" different?

Re:No skill

[Cortexia]

They predicted the games "against the spread", which is very difficult and resulted in 35% win against Vegas. If that required no skill, Vegas would not exist.

Re:The most amazing IN YOUR FACE ever...

[110010001000]

So what? I can pick who is going to the NBA finals most likely. Cleveland and Golden State. You can pick that in the middle of the season. They had the best records and were blowing out teams. That isn't intelligence. I'll go out on a limb and say the Penguins are going to win the Stanley Cup, and I know nothing about hockey or basketball.

Re:Big deal

[wonkey_monkey]

He said you would maximise the number of times you ""win"". He didn't say you could make any money.

How Many AIs got it wrong.

[zam664]

I wonder how many AIs got it wrong and did not get the story. It this is one in 540?

Re: Err, what?

[LordWabbit2]

Why do I get a feeling that you work for Unanimous? But what he is saying is true anyways, whether you like it or not. I know a lot of punters who won't place bets at home etc. but only at the track, where they can get the opinion, or consensus from other punters, in otherwords, a collection of individuals. Why spend two years building an "online platform" when there have been plenty of them built already. Anyone old enough to remember IRC? Hell, you could all just log into World of Warcraft and start "swarming". I would also mention StarCraft, but I think there is too much swarming in that game as it is.

Re:*cough* no.

[sociocapitalist]

It always amazes me how profoundly bad very smart people can be at basic statistics. Granted it's usually not intuitive but it can be learned.

Probably...

Bro, Do You Even Bet?

[Anonymous Coward]

Based on your "common sense" assertion, everyone should win at the track every time by simply using common sense.

As it turns out, most people lose, most of the time. Why do you suppose that is?

Superperfecta in a 20 horse race is 116,000:1 It's very rare that anyone wins the superperfecta. But, by your assessment, it's common sense to be able to pick the winner. I'm guessing you're a very wealthy man?

Re:Bro, Do You Even Bet?

[rmdingler]

Yes. Even if there's something to this swarm intelligence and UNU's ability to reckon it, horse race betting is self-correcting.

If they get the Preakness correct, there will be so many bets on their prediction that the payoff will drop precipitously.

Re:Bro, Do You Even Bet?

[bws111]

The odds of winning the Superfecta are certainly not 116,000:1. If that were the actual odds anyone would be insane to place such a bet when the payout is only $542. Those odds would be true ONLY if it were equally likely for every hose to finish in any given position. Clearly that is not true, or this AI would not have 'picked' a winner, it would just have gotten lucky with a RNG.

I don't know what the Superfecta pool was, but it was probably at least a half a million, so there are probably around 1000 other winners just for this race.

This program picked ONE horse race winner (so far). A whole lot of people can pick ONE winner (especially when picking favorites, which is what this did), but it is rare to be able to CONSISTENTLY do that, which is why most people lose.

Re:Bro, Do You Even Bet?

[rockout]

Everyone is missing the fact that the top four horses to finish were 2-1, 5-1, 10-1, and 12-1, respectively, before the race. They were the four lowest-payout (or likeliest to win, however you want to look at it) horses in the field. Yes, you could pick the top four favorites in every race to finish 1, 2, 3, 4 in that order every time (and I suspect this fake AI would do exactly that, a lot). The crazy thing is that in this Kentucky Derby, that's exactly how the top four came in. That's extremely rare.

Also, 540-1 for a superfecta payout is really low - again, due to the fact that the four favorites came in, in order. Make this same bet on every race and over time, you would lose money just like everybody else.

Re:No skill

[ultranova]

I realize that you are probably associated with these guys and want to hype it up, but I will tell you this: if they had such a method to do this they wouldn't tell everyone about it. They would be making trillions in gambling and the stock market and wouldn't be looking for VC funding.

If you make trillions in gambling, on stock market or anywhere else, you'll attract lots of unwanted attention and likely disappear. Anything that gives you that much advantage makes you a threat to the system itself, so anyone with a stake in it will want you gone. Seeking VC funding is a way of cashing out without getting killed.

Also, how do you propose utilizing a swarm-based anything without gathering a swarm?

Stop the hype.

They can't. People are going to flock to any even remotely plausible-sounding get-rich-quick scheme out of sheer desperation. One might wonder what it tells about the system that everyone's dearest dream is getting free from it, but that would make one a communist (which is bad), or at least that's what Americans have been indoctrinated to believe for the past century or so.

The problem will solve itself either through misery breaking the indoctrination, technological progress making capital as a productive asset irrelevant, or continued wealth concentration draining public's buying power to the point where it can't keep economy going anymore. The last one is currently happening, but I suppose it's still possible a random event might trigger one the first two before the point of no return is reached.