Is Machine Learning the Real Deal or Just a Bunch of Voodoo?
Ever since the great sci-fi writer Robert Heinlein, people have dreamed of machines that could learn and exhibit true intelligence. In the real world, Alan Turing conceptualized it in a paper called Computing Machinery and Intelligence. In practice, however, it took almost another three-quarters of a century to even begin real headway toward machine learning. What is machine learning? Is it real? How is it in use today? What’s the future of machine learning? Here you go.
What Machine Learning Is
The “smart car” of the future won’t just be one that saves you gas money. It will also be able to take over driving duties for you.
Finding a definition for machine learning isn’t easy, and few sources agree on the precise meaning. It’s safe to say that machine learning is something like statistics plus pattern recognition plus artificial intelligence. With this definition, it’s easy to see why we had to first develop the capabilities to collect, store, and analyze huge sets of data in order to achieve machine learning. Without the statistics to analyze for patterns, it is impossible to achieve anything akin to artificial intelligence (AI).
The most obvious use for machine learning, or AI, is the translation of languages. Surely, an intelligent machine could learn any number of languages and translate them with ease in an instant. In practice, this has proven more difficult than expected. Human language is tremendously complex, with limitless idiosyncrasies and oddities. Try translating a page from Chinese or Russian using Google Translate. You can kind of maybe sort of tell what the writer is trying to say. But it is nothing close to the fluidity and accuracy of a human translator.
Real-World Examples of Machine Learning in Practice
It won’t be long before computers are able to make faster, more accurate medical diagnosis than doctors.
Though we haven’t yet fully nailed language translation, there are numerous ways that machine learning is already benefiting society. The first two groups that jumped on the bandwagon were finance (they will literally do anything to increase their earning potential) and security (desperate for ways to stop the ever-growing means by which hackers create mayhem). The finance industry wants to use it to predict future market trends, and security experts use it for detecting fraud, such as credit card fraud.
Earlier this year, a car drove itself across the United States using AI. The autopilot on aircraft, Siri’s (pathetic attempts to) finding information, and Facebook or Twitter recommending people for you to follow are all examples of machine learning. When Amazon recommends products you might like or Facebook recognizes faces in the photos you’re tagging, this is based on machine learning, too.
Your spam filter on your email account and websites that show you ads based on stuff you’ve searched for or shown an interest in — these are also real-world examples of machine learning.
In the future, perhaps the only inhibitor for AI will be our ability to be creative in finding ways to use it. Some practical applications that are underway or coming soon include the ability to accurately diagnose a medical condition based on a patient’s medical history and symptoms. Also, it won’t be long before we have our own self-driving cars (whether we want those or not), and a host of other great innovations based on AI.
Machine Learning as a Service
Who is leveraging machine learning? Well, you can. You can use Syncsort’s data integration tools to collect, prepare and transform enterprise-wide machine data to populate a data warehouse and then leverage Hadoop to run the analysis necessary for AI. There are also services like MLaaS (Machine Learning as a Service) that help turn data analysis into actual machine learning in order to accomplish specific goals.