Dr. Kira Radinsky can see into the future. She is pioneering predictive data-mining software that can forecast natural disasters, disease epidemics, social unrest, and violence outbreaks.
Dr. Kira Radinsky can see into the future. The girl who landed a spot on MIT’s prestigious 35 Innovators Under 35 list this year—previous winners include nerds like Facebook’s Mark Zuckerberg—has figured out a way to forecast natural disasters, disease epidemics, social unrest, and violence outbreaks. But she's no Miss Cleo. Her predictions aren’t vague or ambiguous. They are made of something much more concrete—science. Kira is pioneering predictive data-mining software for Technion-Israel Institute of Technology.
With computers and data mining systems making it easier to see the patterns in the cycles of history, Kira and Microsoft Research co-Director Eric Horvitz believe it’s time to learn from them. The duo developed software that parses the web and composes a complex algorithm that taps into 22 years worth of archives from the New York Times and more than 90 other data sources with an accuracy of 70 to 90 percent. Basically, they're analyzing today’s and yesterday’s news to predict tomorrow’s.
“It’s a very sophisticated form of data mining, enabling deep analysis of disparate events and seeing how they repeat themselves time after time,” said David Shamah of the Times of Israel.
In 2012—after learning that a flood closely following a drought was a preliminary sign of an outbreak of cholera—Kira predicted the first cholera outbreak in over 130 years, something the government and medical experts didn’t see coming.
According to Kira, the system bases its hypotheses on the matches from various algorithms such as “African country with certain level of GDP X, cities with population of Y, and geographical features Z.”
Migrating to Israel as a child from Kiev, Ukraine, she was granted a full scholarship to Technion when she was just 15 years old. Shortly after she went on to win the coveted Israel Defense Prize, interned at Microsoft, earned a black belt in karate, learned salsa dancing, acquired 10 patents, launched the start-up SalesPredict, and completed her PhD—all by the age of 27, earning her academic recognition from tech giants like Google, Yahoo, and Facebook. Perhaps the “27 Club” curse in Kira’s case is bewitching her with brilliance?
Networking with genocide prevention organizations and medical facilities, Kira helps in providing potential solutions for suffering populations and hopes to one day make a “descendent system available to international aid agencies and to the public in general”
I caught up with Kira to chat about her mission to change the world.
VICE: Is it possible to predict the future with today’s technology?
Kira Radinsky: We have reached a critical amount of data and computation power to start finding repeating patterns in history systematically. We built a predictive model based on more than 150 years of historical news data that examines past events with similar outcomes. Our system also incorporates related contextual information pulled from LinkedData, a project that finds connections between hundreds of resources. The combination allows the software to extrapolate from news of a cholera outbreak in Angola, for example, to predict a similar outbreak in Rwanda.
So do you believe that history has a tendency to repeat itself?
The probabilities are always changing, but some patterns, if we abstract them correctly, always remain. And if we incorporate the most recent information we can learn about new patterns emerging all the time. Think about how children learn—they receive reinforcement from the environment and learn patterns. This is also how we learn. I would say the work I have done is not about predicting the future, it is more about making deep analysis on probabilities of future outcomes based on what we have seen, just as an expert in the field would do if he had the time to look at all the available data in the world.
What spurred this passion to use your computer science capabilities to help people?
I became fascinated with the idea of predicting the future at very early age. At some point I came to realize there is so much untapped data that can be leveraged in amazing ways. I believe that not often a person has the opportunity to do something really big that can help many people. My passion is to make big things that can affect people's lives for the better.
With all the focus on the "end of the world," do you see this software being able to predict an end of the world-scale catastrophe like a nuclear holocaust or zombie apocalypse?
The system is built upon probabilities based on patterns it saw in the past. As far as I know, there haven’t been any zombie apocalypses. So far, the system predicted the first Cholera outbreak in Cuba in 130 years, riots in Turkey and Syria, and recently the ones in Sudan. Many critical decisions are not based on real data, because we didn’t have the means to do this. However, using our software, we can now potentially empower important decision makers in the world with the tools to make better decisions.
Do your methods work for looking into the past as well?
In our first experiments, we utilized 150 years of New York Times data and we already have access to data going back to 1500. The system has no limitation on how far in the future it predicts as long as we have enough data to see the patterns. Some patterns have 20 years of distance between the events.
However, because we only have 150 years of data so far, we are limited to these prediction horizons at the moment.
Can this software be applied in other industries?
The implications of this technology also extend to the business world. My current venture, SalesPredict, pioneers predictive modeling to increase sales. We help companies increase their sales pipelines by 75 percent. The system is still piloting and we only now started working with international organizations. However, the product is already working and companies are enjoying it. I believe that this type of technology will be widespread in a matter of years.The predictive algorithms are general purpose, in the sense they can be used to predict a variety of events given the right data. In the future, I believe predictive analytics can be incorporated into everything we do—including predicting mental and physical diseases based on our search behavior.”
More tech stuff: