The new website advertises the power of Microsoft’s Azure machine-learning system to make inferences and predictions.
Microsoft unveiled a website that estimates the ages of people in photographs at its Build 2015 conference this pat Thursday. The website asks users to upload photos, and then in a minute or so delivers a guess for the ages of the pictured people. The website is intended as a showcase for the capabilities of Azure, Microsoft’s cloud-based machine learning system.
Azure uses large data sets to make inferences and predictions based on predictive models. This kind of data analysis very processing power-intensive, and thus has not been put to commercial use. Microsoft is hoping that the website How Old Do I Look will inspire companies to adopt machine learning to help them solve problems.
Right now, the website is not 100 percent accurate. It guessed that 47-year-old Microsoft CEO Satya Nadella was 59 based on a low-resolution picture. The website’s predictions should become more accurate with time, however, because machine learning’s models get better as more data becomes available to analyze.
How Old Do I Look can also be used to estimate the age of figure in paintings. The website guessed the Mona Lisa’s age much more accurately than it did Mr. Nadella’s, estimating the model’s age to be 23.
Azure, which launched in February, is already being used some real-life corporate applications. The Japanese company Fujitsu has used the system to help predict the best times to fertilize dairy cows to insure a high rate of pregnancy.
Azure is able to predict the best time to fertilize cattle by analyzing data on how many steps each cow is taking Cows in heat tend to walk around more than normal, and Azure tells farmers which cows are unusually mobile. Azure’s predictions for which cows will become pregnant is 95 percent accurate, which is almost three times as accurate as the previous method of watching out for which cows were walking more than others.