Universal Learner algorithms taking over the world
Yesterday I had a chat with video rental customer service because my rental movie suddenly stopped downloading. Mark from support very politely helped me. About 10 minutes later, my problem was solved, and I was able to watch the movie. ”Excellent service” commented my wife as the movie night finally started.
Yes, Mark surely saved movie night this time, and I felt that I had received great service from a nice person—except I would bet Mark was not a person. I think Mark was a quite advanced customer service robot that very smoothly adapted to my language and guided me through a preset protocol to fix the problem with my download. Mark even processed a refund to my account and pleasantly said goodbye at the end of the conversation.
Today, there are many examples like this in the digital universe. Sport and stock market news are generated by dedicated algorithms. Indeed, algorithmic stock trading has become commonplace. We all remember how IBM’s Watson supercomputer beat the top human Jeopardy! contestants in January 2011. More recently, Google’s DeepMind defeated a world champion of the game Go, which is exponentially more complex than Chess. (By the way, IBM’s Deep Blue already beat Chess grandmaster Garry Kasparov in 1997.)
Machine learning is fascinating. Over the decades, sci-fi literature and films have portrayed a world filled with machines that are more intelligent than humans. Ray Kurzweil called this turning point the Technological Singularity and famously predicted it would happen around the year 2045.
Today, we are still far from the Singularity. But exponential technologies like computing power, network bandwidth and storage together with cutting-edge algorithms have enabled artificial intelligence to develop rapidly. Now we see new applications of machine learning almost every day. Algorithms can make accurate diagnoses by utilizing huge online patient databases. But Sentrian, a California-based technology company, takes computer assisted diagnosis even further. By using wireless biosensors, they are able to monitor and collect basic data like blood pressure or heart rate as well as blood potassium levels. With constant monitoring of these types of parameters the algorithm can spot even the tiniest changes and thus predict severe conditions in advance. For example, this type of monitoring can predict heart failure up to ten days in advance.
Today's algorithms are domain specific, yet very sophisticated. But when programming these algorithms, designers put a lot of domain knowledge into the algorithms to ensure they work properly and efficiently. Accordingly, such algorithms cannot operate beyond the environments for which they were designed. To reach the Technological Singularity, researchers must solve this problem. This is why so much effort has been dedicated to developing a Universal Learner (UL) algorithm that can figure out its environment and adapt to it. This is the key to the true artificial intelligence. If a computer program can figure out what can or should be done in any given environment and produce new knowledge, it would change the world dramatically. For example, an algorithm could invent new algorithms. Computers would thus create new unique knowledge. Some white-collar jobs would disappear from societies because humans are too slow to learn. Computers would begin to drive cars, fly aeroplanes and eventually controlall vehicles. This change would represent a fourth Industrial Revolution.
The possibilities are limitless. UL algorithms could examine the history of mankind and create powerful new knowledge. This could be applied everywhere. UL might revolutionize fashion by designing the next summer collection. UL could be the next big media producer replacing social media. Societies and businesses would utilize UL to replace jobs consisting of repetitive tasks in order to become more efficient and productive. UL may even someday produce cancer cures or ways to beat viruses. Finally, we humans may even augment ourselves with UL to become smarter, wiser and ultimately happier.
The fuel for the next industrial revolution is-data, and the engine is Universal Learner algorithms. This engine will process our ever-growing zettabytes of data and makes sense of everything. In the future, Universal Learner technology will help us to understand the world and improve our lives. I think that is a goal worth aiming for.