How Machine Learning Shapes Artificial Intelligence Technologies

    Artificial intelligence (AI) is actually a group of linked technology. It is complex and, to many people, a bit frightening. It is also huge and expanding rapidly.

    That size is evident in both value projections and the expansive range of industries using the technologies. In February, Markets and Markets valued the 2017 AI market at $16.06 billion. That number is expected to reach $190.61 billion by 2025. That’s a compound annual growth rate (CAGR) of 36.61 percent. The broad range of industries is similarly impressive. The firm said that AI will be used in health care, manufacturing, automotive, agriculture, retail, security, human resources, marketing, law and the financial technology segments.

    The bottom line is that AI is here today. For instance, the suggestions made by a streaming service’s search engine or the ads selected by a website to display on an individual’s computer ever more accurately portray his or her tastes. This is a use of AI. More specifically, it is an example of machine learning, which is one of the building blocks of AI.

    Machine learning combines with at least two other technologies to form AI. Natural language processing, as the name implies, is the science of understanding verbal communications. This will extend beyond decoding words and include nuance and emotion. For instance, somebody saying, “Yeah, right” to a question may be facetious (especially if that person is a teen) or really mean it as an affirmative response. An NLP engine of the future would be able to decide between those possibilities.

    The second is computer vision, which entails more than being able to decipher images or video. For instance, a human generally has little difficulty identifying a person’s image though years have passed, his or her face has aged and their hair style has changed or disappeared. That is a challenge for the computer vision, however, said Kimberly Nevala, the director of Business Strategy for SAS.

    The Role of Machine Learning

    Thus, there are three main pieces to the AI puzzle. They may be combined differently to achieve desired goals. A specific use case may require other elements to add to this core.

    Machine learning is the glitziest element of AI and perhaps the element that scares people. “Machine learning is both a noun and a verb,” said Michael Wu, Ph.D. and Chief AI strategist for PROS, a company that uses AI and other tools to provide dynamic pricing to its clients. “As a verb, machine learning is the process of turning data into models. [It can help] predict the future and do a lot of different things.”

    Typically, a computer operates on the instructions given to it by humans. That becomes more and more inefficient as the tasks it is being called upon to do grow more complex, the original assumptions upon which those instructions were based change, and the very goals of the organization evolve. In machine learning scenarios, the AI system takes the input it gathers and adjusts its mission along the way.

    Here is an example of how these elements can work in concert: Suppose an AI platform is used to provide information to a cable television operator’s call center representative based upon the questions a customer asks. In the future, the NLP element of the AI platform will be able to assess the emotional status of the person calling the contact center by her voice: Is she annoyed? Likely to churn? Or, conversely, is this a good time to try to upsell the person? Based on that, the optimal tools are provided to the contact center employee. In the long run, the machine learning element of the AI platform will use that experience to better tailor the guidance and materials it provides to the contact center agent in the future.

    Another example focuses more on the machine learning element. There are, of course, only four basic directions in which a vehicle can travel: forward, backward, left and right. It is impossible to program an autonomous vehicle (AV) on how to react in every instance (“If there is a dog on the right, move to the left,” etc.). Computer vision enables the AV to identify the obstacles and set in motion the appropriate reaction. The machine learning element will store that knowledge (“If a dog is on the right, move left”). “There are massive combinatorial problems,” Nevala said. “Actions are well defined, but how to use them is complicated. That is where machine learning comes into play. It enables [the vehicle] to learn from experience.”

    The Two MLs

    There are two types of machine learning, supervised and unsupervised, said Vince Jeffs, the senior director of Product Strategy at Pegasystems. He said that the majority of machine learning in use is supervised.

    Supervised machine learning occurs in instances in which the computer is given input and makes a binary choice (a yes or a no). If the machine is given more information – or “training” – until it gets the answer right, it subsequently has the ability to make the right decisions in similar situations.

    Unsupervised learning, as the name implies, provides the AI platform with a large amount of data and lets it essentially figure out what to do with it. “It is trying to make sense of its world by putting things into buckets,” Jeffs said. “It is an approach in which we are not giving it any instructions. We are not ‘leading the witness,’ as it were.”

    Machine learning is a never-ending process. Suppose, for instance, a machine learning algorithm and the AI platform with which it is associated suggest that a company with three employees in the shipping department should hire two more. The subsequent data is continually assessed. The conclusion may change over time and new data enters the system. It may be found, for instance, that five people is too many and one should be reassigned. Or that five is not enough because adding the extra two resulted in so much new business that a sixth now is necessary. The bottom line is that it is a continual process.

    Carl Weinschenk covers telecom for IT Business Edge. He writes about wireless technology, disaster recovery/business continuity, cellular services, the Internet of Things, machine-to-machine communications and other emerging technologies and platforms. He also covers net neutrality and related regulatory issues. Weinschenk has written about the phone companies, cable operators and related companies for decades and is senior editor of Broadband Technology Report. He can be reached at and via twitter at @DailyMusicBrk.


    Carl Weinschenk
    Carl Weinschenk
    Carl Weinschenk Carl Weinschenk Carl Weinschenk is a long-time IT and telecom journalist. His coverage areas include the IoT, artificial intelligence, artificial intelligence, drones, 3D printing LTE and 5G, SDN, NFV, net neutrality, municipal broadband, unified communications and business continuity/disaster recovery. Weinschenk has written about wireless and phone companies, cable operators and their vendor ecosystems. He also has written about alternative energy and runs a website, The Daily Music Break, as a hobby.

    Get the Free Newsletter!

    Subscribe to Daily Tech Insider for top news, trends, and analysis.

    Latest Articles