Sensorization and machine learning

All the signs indicate that we are headed toward a future in which many aspects of our lives will be monitored by sensors. The earphones in the photograph are the latest from SMS Audio, the company created by US rapper 50 Cent and based on Intel technology. They are designed to monitor physiological variables associated with physical exercise, and seem like a more comfortable alternative to wearing an armband, bracelet, or watch when practicing sport.

But the earphones are just a small part of the huge jigsaw puzzle behind many of the recent developments in the technology sector: yesterday Samsung announced that it had bought SmartThings for $200 million, positioning the South Korean giant in the home automation sector, and converting the founders of a company that got going thanks to Kickstarter into millionaires. We seem happy to move toward an increasingly monitorized existence, even though we have no idea who will take the blame when the information gathered by one of these sensors leads to a mistake.

Smartwatches, bracelets for monitoring the elderly, new longer-life batteries to power them, and an avalanche of information on every step or breath we take. Data of all kind, with any number of good and bad uses, are going to change the rules of business, and that are even threatening international agreements.

And what are we supposed to do with all this information. We are already saturated, and that is by just analyzing around 1% of data. The logical thing to do, in fact, virtually the only thing we can do, is… get other machines to analyze the data. Machine learning is the next great frontier, and the only way that gathering so much information can possibly make sense. A pilot program using an algorithm to analyze the data of 133,000 patients from four hospitals in Chicago between 2006 and 2011 predicted emergency situations four hours before doctors did.

Revising a patient’s medical history, combined with data on age and family antecedents, after being analyzed by an algorithm can drastically reduce deaths in situations where rapid medical assistance is vital.

Monitoring our health parameters will become easier, cheaper, and more accessible, but it will need to be matched by tools to analyze the information gathered. The potential of this market is huge, and we can expect some major events soon.

(En español, aquí)