Scientists at the Massachusetts Institute of Technology (MIT) have created an artificial intelligence system called Data Science Machine that can replace human intuition with algorithms in the big-data analysis. Computers are known to be great at crunching numbers, but it requires some human intuition to choose which “features” of the unfamiliar data to analyze. The Data Science Machine developed by MIT researchers not only searches for patterns but also designs the feature set, taking the human element out of the big-data analysis.
MIT’s AI system beats 615 of 906 human teams
Until now, humans have been more efficient at filtering out decisions that don’t make sense, even though it is a possibility, from gigantic pools of figures. But MIT’s AI system has outperformed its human counterparts in human intuition. MIT scientists enrolled the Data Science Machine in three competitions to test the first prototype of their system. The machine was competing against a total of 906 human teams to find predictive patterns in unfamiliar piles of data.
The Data Science Machine performed better than 615 of 906 teams. It had an accuracy rate of 94% and 96% in the first two competitions, but fell to a “more modest” 87% in the third competition. The machine demonstrated its capabilities by performing better than even the smartest people when big-data analysis is concerned.
The Data Science Machine is more efficient than humans
What’s more, MIT pointed out that its AI system is far more efficient than humans. It usually takes humans several months to choose which data sets should be analyzed. In contrast, the Data Science Machine takes just two to 12 hours to produce each of its entries. Max Kanter and his thesis adviser Kalyan Veeramachaneni will present the study at the IEEE Data Science and Advanced Analytics conference in Paris later this week.
Researchers applied the algorithms to numerous practical problems such as predicting which students are at the risk of dropping out of online courses, and determining the power generation capacity of wind-farm sites.