the magazine that fast-forwards your career.
Here's how to read the magazine:
by clicking the arrows at the side of the page, or by using the toolbar.
by clicking anywhere on the page.
by dragging the page around when zoomed in.
by clicking anywhere on the page when zoomed in.
websites or send emails by clicking on hyperlinks.
Email this page to a friend
Search this issue
Index - jump to page or section
Archive - view past issues
Real Business : 2013 Issue 1
of things they wanted to do with their data, but I’d ask ‘how far along are you in terms of implementation?’ and the enthusiasm for data wasn’t really matched by the result,” Goldbloom says. “I realised part of the problem was that you have this market failure – there are companies who wanted to hire data people without really knowing where to start. But I knew if I knocked on their door and said ‘hey, let me at it’, they’d say, ‘what makes you qualified to solve this problem?’.” It was out of this sense of market failure that the idea of Kaggle was born. Goldbloom admits its first competition was somewhat frivolous, with analysts asked to predict the next winner of the Eurovision Song Contest. However, after getting closer to the final result than the bookmakers, Kaggle earned media attention and a much more serious challenge – predicting the progression of HIV cases from a medical database. The results were astonishing. “We were able to outperform four years’ worth of scientific research in just three months,” Goldbloom says. “That result was enough to be able to go around and say, ‘this is a serious business. I can do some seriously powerfully work with this platform’.” WiTHiN THrEE YEArS Kaggle secured US$11 million in venture capital funding from some of Silicon Valley’s highest-profile entrepreneurs, such as PayPal co- founder Max Levchin, Google chief economist Hal Varian and Silicon Valley icon Ron Conway. While Kaggle’s public competitions have captured the headlines, it’s the company’s private, invitation-only contests that generate the bulk of its revenue. The company makes its money by charging fees to companies that r un private competitions through the site, including fees for preparing data, designing problems and selecting experts to participate. Goldbloom says the commercial potential is significant. “Just about every company needs to predict something, whether it is a bank predicting who’s going to default on a loan or an insurance company predicting who’s going to crash their car.” Kaggle’s business model harnesses the power of crowdsourcing the world’s best data scientists to take advantage of the information. Each solution is tested against historical data, with competitors given rapid feedback on the accuracy of their predictions and the results posted on a leaderboard for all competitors to see. “ When you can objectively measure how well somebody’s performed and, even better, show their performance on the leaderboard in real time, it’s incredibly powerful,” Goldbloom says. And the rewards can be compelling, particularly for those chosen to work on private projects. “ We’ll see the world’s best data scientists earning what the best hedge fund managers or the best golfers make, and the flipside is also generating an enormous amount of value for companies.” KAgglE iS NoW based in Silicon Valley, Califronia, with Goldbloom citing better opportunities and a more developed understanding of tech start-ups as the main reasons for the move. “It became clear that, while in Australia it was a constant battle to explain what predictive modelling was, let alone why crowdsourcing was a good approach to it, here you had all these people and companies that just got it instantly,” he says. Goldbloom regrets that there are relatively few opportunities for Australians to be exposed to the inner workings of a start-up. “Success begets success and it would be really nice to get to that point with Australian companies as well.” For now, Kaggle is focused on growth, using its new capital to boost its team of developers and in-house data scientists, as well as hiring people to help with sales. The pool of data scientists available to compete on Kaggle projects has also grown exponentially. “Just under a year ago we had 10,000 to 11,000, now we ’re almost 46,000 – so that kind of gives you a rough sense of the order of magnitude by which Kaggle’s grown,” Goldbloom says. He is also focused on forming strong relationships with companies seeking to make the most of the data they own. “ We’d like to see most of the world’s companies relying on Kaggle for their most valuable predictive modelling problems.” Kaggle founder Anthony Goldbloom is helping companies make the most of their data. 27 Predicting the winner of the 2010 World Cup in the lead-up to football’s World Cup, Kaggle invited its members to take on analysts at major investment banks in forecasting the results. sixty-five teams participated, with the best- placed banking team finishing 28th. Betting markets finished in 16th place behind the Kaggle member teams. The winner was australian economist Thomas Mahony. helping nAsA map dark matter a 2011 competition hosted by nasa aimed to help with the mapping of dark matter in the universe, based on images from more than 100,000 galaxies, blurred to simulate the distorting effects of dark matter. Within a week, British glaciologist Martin O’leary submitted the winning entry. The White House announced that he had “outperformed the state-of-the- art algorithms most commonly used in astronomy for mapping dark matter” and his solution has provided a new way of mapping galaxies millions of light years from earth. reducing costs for marking student essays Goldbloom says that one of the projects that most blew him away involved looking at more than 24,000 essays from students around the us, each of which had been graded by two teachers. The purpose was to see if an algorithm could grade an essay with the same level of reliability as a human teacher. They discovered that the winning algorithm was about as reliable as a teacher, with a similar level of variability between different markers. While there are limitations to using algorithms for marking, the results open up new possibilities for efficiently and cost-effectively assessing student performance. KAGGle ComPeTiTions
Issue 3 2012