Technologies that employ artificial intelligence are all around us in our daily lives. AI is not just what’s seen in science fiction movies or lauded by tech giants claiming to be stepping closer to that of human intelligence. From Apple’s Siri and Amazon’s Alexa, to IBM’s Watson and supercomputers, to simply the personal computers prevalent in everyday life, they are all examples of artificial intelligence.
On a fundamental level, AI systems generally demonstrate at least one of the following characteristics of human intelligence: knowledge, perception, motion, manipulation, planning, learning, reasoning, and problem solving.
AI can be further split into two categories: narrow and general AI. The latter is adaptable intellect as found in humans (AGI), a step away from the “superintelligence” found in the likes of science fiction where robots are taking over the world. One of the classic, modern examples of AGI is IBM’s Watson beating human contestants on the trivia game show, Jeopardy.
This branch of AI will take several more decades to fully accomplish and even more to reach a level where it becomes more mainstream. According to ZD Net, the likelihood of so-called “superintelligence” occurring will increase to 90% by 2075. By then, Siri and Alexa will have upgraded to true, fully functional personal assistants. Whether AI will get to the point of Stephen Hawking’s “singularity” is a whole other discussion.
Currently, Siri and Alexa fall into the category of narrow AI as they only recognize speech and language and search information based on those perceived words. Narrow AI are the computers around us which carry out specific tasks without being explicitly programmed to do so. These systems can only learn or be taught to do certain tasks while general AI is being able to learn vastly different tasks or reason based on experience. The scope of actions performed by AGI is unlimited while narrow AI is limited to its code and algorithms. It cannot perform other types of tasks outside of the ones it’s been programmed to do.
Within AI are Neural Networks, Machine Learning, and Deep Learning. AI roughly developed in this order over the decades. Neural Networks are the structures of how basic AI works. They networks of interconnected layers of algorithm inspired by the human brain. Aptly called neurons, these networks feed data into one another at different inputs until the output is close to what is desired. At this point, the system has “learned” the particular task.
Machine Learning is a subset of Deep Learning where neural networks are further expanded into much larger and numerous interconnected networks and layers. The AI is trained using massive amounts of data to learn particular tasks. For instance, an AI is given thousands of photos in order to recognize a human face. The AI infers what it is being given, and validation is done afterwards to test its learning. Machine Learning can be either supervised, where data is annotated for features of interest, or unsupervised where algorithms attempt to identify patterns in data that can be used to categorize future data. Many data analytics tools use this process called classification.
Today’s AIs are in the stage of deep learning, and its evolution and resurgence into the limelight is fueled by the easy availability of large amounts of data and parallel computing. Artificial intelligence automates repetitive learning and discovery. Besides being potential sources of human assistance such as through the development of self-driving cars, AI allows businesses to gain insights from big data generated everyday with more accuracy, speed, and depth.
AI is the only feasible and cost-efficient method of managing and analyzing data as technologies generate greater and greater quantities of data from all sorts of sources and devices. Soon, Artificial Intelligence will no longer be on the backburner of everyday people’s minds. It’ll be at the forefront, paving the way for more streamlined businesses and comfortable lives.
Oct. 25, 2019