The Machines are Learning
Imagine for a moment if you could see the billions upon billions of data points flowing throughout our world. As our society has embraced more digital ways of being, we generate vast amounts of data. Each one of us, as we engage with digital tools, lay down trails of data that say something about events, places, and who we are. Within the mysterious realm of Artificial Intelligence (AI) lies the magic of Machine Learning of ML, a science of data and algorithms that act on it.
Riders of buses in the Greater Vancouver area were frustrated that Google Maps, and other apps relying on Translink’s online data feed, predicted earlier arrival and departures by over 5-minutes the majority of the time. They were often standing in the rain or dark longer than they would like waiting for a bus that from their perspective was late. By partnering with Microsoft, Translink was able to bring together numerous historical and real time data feeds including traffic and weather data, and develop a machine learning algorithm to improve predictions by 75%. Annually Translink experienced 252 million bus trips over 200 routes with 1500 buses. There new AI made 20000 predictions per day. This is the power of cloud computing scale machine learning delivering goodness to people.
In the Seattle area, there are numerous traffickers in the sex slave business. Victims are lured into slavery by being told no one cares and there is no hope. Seattle Against Slavery built Freedom Signal Victim Outreach and Buyer Deterrence, an AI solution in Microsoft’s Azure cloud. This system indexes online ads, reviews pictures, lifts phone #s, victim age, and other information enabling advocates to send text messages of hope and help to hundreds of potential victims, letting them know there is a way out. Law enforcement use Intercept ChatBots to post decoy trafficking ads online and wait for buyers to respond. The bots, built with Language Understanding, a component of Azure Cognitive Services, converse with prospective sex slave buyers, probing their intentions and ultimately disrupting upwards of 10000 searches per day to buy sex with children!
It is now possible to gain reasonably accurate understanding of how people feel about a brand, an event, a person, etc. This example demonstrates how a sentiment analysis tool, another machine learning example, scours Facebook and Twitter posts about Uber and presents evidence of how people feel about Uber from a price, safety, service, and other factors. This is a valuable tool for a brand. This would be a valuable tool for teachers to use in analyzing student work or employers to analyze how their employees feel about their workplace, etc. But, this could potentially be weaponized by political campaigns and used to inject material into the social media stream to influence and change people’s sentiments towards a party or candidate. With Microsoft into the Sentiment Analysis game as a cloud service (here), we will see an developers use sentiment analysis to create interesting apps in the future. Bots may become indistinguishable from humans – this could certainly become problematic as well as provide useful services.
I hope the future uses of machine learning systems tend to be more for good than bad!