The year 2015 will be remembered for various tech developments, particularly with regard to the Internet of Things (IoT) and big data. Along with these two developments, researchers in the fields of artificial intelligence (AI) and computer security are concerned about what lies ahead for 2016, particularly within the realm of machine learning through the mining of big data.
The Big Data and Machine Learning Nexus
Machine learning is a vital component of AI development. In the 20th century, AI researchers saw great potential in terms of computers performing tasks related to data filtering and pattern recognition, two functions that required some degree of learning by means of coded algorithms. By 2001, when Microsoft introduced the groundbreaking Windows XP operating system, machine learning had advanced to the point of becoming mainstream; case in point: software applications dedicated to speech recognition, semantic search and predictive analytics.
Big data has been developing parallel to machine learning in the 21st century. Developers have perfected data collection and storage on a massive scale; now it is up to AI software to do something with all that information. The data processing applications of the 20th century cannot parse the colossal datasets produced these days; thereby, machine learning techniques are being applied to understand how information can shape our world.
To develop and improve smart applications such as Apple Siri, Google Now and Microsoft Cortana, one of the techniques being used is called supervised learning. The speech and image recognition abilities of these smart personal assistants are being honed by machine learning processes that start with massive data sets of audio and image samples. Armed with algorithms and millions of human speech recordings, the supervised learning process directs neural networks to engage in pattern recognition tasks.
Mathematicians who observe AI processes are impressed with the efficiency of sheer efficiency of the big data supervised learning process within neural networks; nonetheless, these same mathematicians are not yet able to explain why this is such a success.
Instead of worrying about the underlying math that is causing AI systems to become smarter, research and development teams are moving in the direction of unsupervised learning. Imagine neural networks running advanced pattern recognition algorithms that can deal with longitudinal surveys such as 300,000 people answering 300 questions that reveal their personal approach to problem solving. Such a task would produce one dimension for each question, and the right algorithm could find hidden data structures that may reveal valuable knowledge for AI processes to learn. At this point, the mathematical explanation is irrelevant but the potential for application is fascinating.
The Need for Security in the Age of Big Data and Machine Learning
Bigger data sets translate into a bigger need for information security. The cloud computing paradigm comes with the caveat of a dark side. A December 2015 report issued by esteemed security firm McAfee Labs indicated that hackers are creating their own massive data sets of stolen information such as credit card data, social security numbers, usernames, and passwords.
Armed with big data, cyber crime outfits are conducting operations similar to those carried out by tech giant IBM in the development of their Watson AI. Hackers are breaching networks and going after major databases so that they can link information and even run pattern recognition algorithms. The idea is to come up with intelligence that will prove more attractive to potential buyers who frequent black markets.
Now that IBM offers access to Watson via application programming interface (API) access, it is more important than even to enact operational security. AI systems are learning and becoming smarter, and the same can be said about cyber criminals who wish to target big data. In 2016, protection of cloud storage systems will be paramount to guarantee continuous learning by AI systems.
