Home » Artificial Intelligence » Machine Learning: Both Deep and Wide

Machine Learning: Both Deep and Wide

June 11, 2014

supplu-chain

Earlier this year, Google paid more than $400 million “to acquire DeepMind Technologies, a startup based in London that had one of the biggest concentrations of researchers anywhere working on deep learning, a relatively new field of artificial intelligence research that aims to achieve tasks like recognizing faces in video or words in human speech.” [“Is Google Cornering the Market on Deep Learning?” by Antonio Regalado, MIT Technology Review, 29 January 2014] According to Regalado, “DeepMind’s expertise is in an area called reinforcement learning, which involves getting computers to learn about the world even from very limited feedback. ‘Imagine if I only told you what grades you got on a test, but didn’t tell you why, or what the answers were,’ says [Yoshua Bengio, an AI researcher at the University of Montreal]. ‘It’s a difficult problem to know how you could do better.'” Regalado goes on to explain that this kind of research could lead to better search engines “and might be particularly useful in helping robots learn to navigate the human world.” Google has been very active in the area of robotics and driverless cars.

 

Ashlee Vance calls deep learning “a funky part of computer science seen as key to building truly intelligent machines.” [“The Race to Buy the Human Brains Behind Deep Learning Machines,” Bloomberg BusinessWeek, 27 January 2014] She goes on to report, “Some members of DeepMind have published papers over the past couple of years through NIPS, the Neural Information Processing Systems foundation, that provide some insights into their work. One describes a set of algorithms that can mine a social network for patterns at record speed.” In the retail world, where trends are king, it’s not difficult to understand why such algorithms are highly desirable. Rick Delgado insists that machine learning applications will have broad appeal in the commercial sector. “Companies around the world both large and small are always looking for the newest, most innovative ways to expand their businesses,” he writes. “One useful and rapidly growing tool is the implementation of machine learning.” [“Making Machine Learning Work for Business,” SmartData Collective, 24 March 2014] He continues:

“Put simply, machine learning means getting computers to perform tasks without having specific programs written for them. It’s more prominent than you may think, and its uses in the business world are just beginning to have a big impact. Machine learning is a type of artificial intelligence that takes vast amounts of data and makes sense of it. Many businesses have started turning to machine learning as a way to sort through their collected data, streamlining their work and improving the customer experience.”

It needs to be pointed out that there is a difference between machine learning and cognitive computing. It’s a difference that is often overlooked or misunderstood. Let me give you an example. Delgado points out that one way machine learning is used today is to detect spam emails. He explains:

“Through machine learning, the program’s algorithm is modifiable to better detect a piece of spam mail. The program receives input from the email user, causing it to adapt to whatever is tagged as spam. With enough input from enough users, the program can then filter out spam mail with greater accuracy.”

The computer “learns” what words are commonly associated with spam messages and filters them out. That’s why spammers deliberately misspell words to try and trick the algorithms; but, machine learning systems soon catch on. Just because certain words trigger filtering actions by the algorithm, it doesn’t mean that the computer understands what those words mean or why it has been instructed to perform certain actions. A cognitive computing machine takes that next step. It understands a word’s meaning in the context it is being used and can make rational decisions about it. In other words, a cognitive system, like the Enterra® Cognitive Reasoning Platform™ (CRP), uses both reasoning and computation to achieve insights. We call this a “Sense, Think, Act, and Learn®” process. Edd Gent reports, “In a white paper [released earlier this year, the] IT professional body BCS highlighted cognitive computing as one of the next great waves of computing, but one that also has the potential to become the most controversial technology in the world by the end of the decade. Using a combination of artificial intelligence and machine learning, cognitive systems continually learn from the data fed into them, refining the way they work, anticipating problems and inferring solutions.” [“The rise of the thinking machines,” Engineering and Technology Magazine, 31 March 2014]

 

Although I believe that cognitive systems represent the future of computing, machine learning remains an important and useful step in the evolution of computing. John M. Boyer states, “Machine learning offers another leap forward in the effectiveness and hence value of machine intelligence.” [“Machine Learning versus Machine Intelligence,” Smarter Everyone, Smarter Everything, Smarter Everywhere, 22 January 2014] How useful is machine learning today? Delgado goes to explain a few more ways machine learning is benefiting businesses today. He writes:

“Businesses are … using machine learning to better connect with customers through advertising and marketing. … Machine learning is able to take the data you have input in relation to the business and better collaborate with different systems and employees to make sure that the messages you receive are both relevant and timely. … Companies have also turned to machine learning to improve the customer experience on your smartphone or tablet. They’ve put machine learning into practice to observe a person’s behavior while using apps, and from there, it can predict that user’s interests. From those predictions, the program can then recommend items the user might want to purchase. … Those same predictive qualities are already being used in other internet businesses. Facebook takes input from millions of people, and through machine learning is able to recommend connections with individuals you might know, or might want to get to know. Netflix takes the same approach, only by finding specific movies and genres that interest you. Google takes your entire web history and activities and is able to tailor the ads you see in the sidebars and on the search engine to products you might be interested in buying. This more personalized approach requires sorting through vast amounts of data.”

Machine learning is also being used in ways that improve the quality of your life or protect your resources. Delgado notes, for example, “Some businesses and organizations are using machine learning to predict future wait times in emergency waiting rooms. These programs take into account a number of variables and data, including staffing levels, emergency room layouts, and patient data.” Machine learning is also used to identify fraudulent banking activity associated with your accounts. In another article, Delgado suggests that you should seriously reconsider using a Big Data analytics vendor that doesn’t offer solutions that utilize machine learning. “The complexity of machine learning and predictive analytics can … drive some businesses away from using them,” he writes, “but that’s why a good big data vendor will actually leverage machine learning as a way to remove that complexity.” [“Why Machine Learning Matters When Choosing a Big Data Vendor,” SmartData Collective, 14 May 2014] He concludes, “One of the most important choices to make is to find [a vendor] that uses machine learning. Without it, analysis of big data simply isn’t as effective, and much of that data could be misinterpreted or mishandled.” The reason for this is that machine learning goes beyond simply analyzing a wide variety of data to diving deep into it and learning from it.

Related Posts: