Over the years, human beings have been teaching machines to act on their own. However, even today, the machines do not make all decisions autonomously. In fact, the data scientists who perform machine learning do not create Artificial Intelligence (AI) in many cases. Instead, they develop basic predictive models. In these instances, the machines do not make automatic decisions; educated, trained experts make the final calls. The human experts use the predictive models as decision-making aids.
Deep learning is more powerful. It enables machines to make autonomous decisions. Some of the popular applications of deep learning are in the areas of automated driving, medical research, aerospace and defense, and industrial automation. Companies have been applying deep learning techniques to create automated vehicles that do not require human supervision. These self-driving vehicles detect stop signs, traffic lights, pedestrians and other vehicles. They monitor critical road conditions to safely travel from one place to another. Medical researchers have been applying deep learning to create machines that can observe patterns and detect various diseases, tumors and cancer. Deep learning is used in aerospace and defense organizations to sift through streams of data in real time, looking for signals or targets of interest. This helps to assess battle scenarios and enable faster situational analysis in the air or on the ground. Deep learning has been helping industries in preventative maintenance and repairs, condition monitoring to improve machine efficiency, and optimizing supply chains.
Krishna C. Mukherjee is one of the key driving forces behind ushering in this fascinating era of AI. He is a former Microsoft executive. His illustrious professional career spans over three decades – from 1988 to present. He played a pivotal role in the architecture, design and development of Microsoft’s most famous products – Office and Windows. In the early 1990s, he introduced AI technology at Microsoft and was at the heart of AI projects undertaken at the company. In the late 1990s, he invented the AI technology called “Intelligent Filing Manager” to automate workflows and make business processes highly efficient. He established the Software as a Service (SaaS) model and paved the way for efficiently performing business transactions on the cloud. He directed the development of the popular Bloomberg Valuation Service, or BVAL, that accurately prices millions of financial instruments across multiple asset classes. He used quantitative algorithms and AI to develop BVAL. Thus, Mukherjee introduced AI into the finance industry and helped to build the foundation of modern financial technology, or FinTech. Subsequently, Mukherjee created the AutoPay platform and revolutionized the payments industry. He has implemented omni-channel retailing applications with built-in recommendation systems. He has been creating AI-based customer services solutions with interactive voice response (IVR) systems and chatbots that have natural language processing (NLP) and machine learning capabilities.
Impact of AI
Man has long feared the rise of the machine – his own creation surpassing him. While AI is rapidly changing the world, humanity does not need to be afraid of AI.
The endless possibilities offered by AI have made it a subject of study for almost everyone. In the summer of 1956, John McCarthy held the first workshop on AI at the Dartmouth College in Hanover, New Hampshire. To understand the importance of the advances in AI, it is necessary to know what AI is today and where it has come since its official birth.
It is undeniable that AI has brought about social changes. However, we still do not know all the implications – good and bad – of these changes. To complicate matters, many in the media talk about AI and assume changes that have not yet happened. These types of talks tend to confuse the audience, who are interested to know the real impact of AI and not the exaggerations surrounding it. Therefore, it must be possible to dispel misinformation and separate the facts from the myths to gain a true appreciation of AI.
Machine learning and deep learning
Understanding the latest advancements in AI can prove to be overwhelming. Many of the AI innovations can be narrowed down to the two concepts – machine learning and deep learning – that we have looked at earlier in this article.
Zendesk defines machine learning as “Algorithms that parse data, learn from that data, and then apply what they have learned to make informed decisions.” Essentially, machine learning is a tool for converting information into knowledge. In recent years, there has been an explosion of data from the Web, e-mail messages, social media content (tweets, blogs, posts on Facebook), as well as machine-generated data from sensors and from electronic trading systems. As the Internet of Things (IoT) becomes more prevalent, the volume of data is going to grow even more. For all this data to be useful, we need to analyze and find the patterns hidden within it. Machine learning techniques are used to automatically find these underlying patterns that we would otherwise may not be able to discover.
Krishna C. Mukherjee has been building predictive recommendation systems. A simple example of a machine learning algorithm is an on-demand music streaming service that can make decisions about which new artists or songs to recommend to a listener. Such machine learning algorithms matches the profiles of listeners, associate a listener’s preferences with other listeners who have a similar musical taste, and make recommendations. This matching technique is used in types of other services that offer automated recommendations. Online stores, for example, use recommendation systems to suggest ensembles of associated products to the shoppers. The predictive search feature of search engines predicts a user’s search query as it is typed. This feature provides a dropdown list of suggestions that changes as the user adds more characters to the search input.
Traditional software engineering combines human created rules with data to find the answers to a problem. Machine learning, on the other hand, uses data and answers to discover the rules behind a problem. To learn the rules governing a phenomenon, machines have to go through a learning process, in which they try different rules and learn from how well the rules perform. There are multiple forms of machine learning: supervised, unsupervised , semi-supervised and reinforcement learning. These forms have varying approaches, but they follow the same underlying theory and process. A lot of times, the lines between these different forms of machine learning blur.
Krishna C. Mukherjee has been exploring deep learning approaches to build conversational AI systems. Deep learning can be considered to be a subset of machine learning. However, its capabilities are different. While machine learning models do become progressively better at performing their tasks, they still need human guidance from time to time. If a machine learning algorithm returns an inaccurate prediction, then a software engineer has to adjust the algorithm. With a deep learning model, an algorithm can determine on its own if a prediction is accurate or not, without any human guidance. A deep learning model is designed to continually analyze data with a logic structure that is similar to how a human brain would draw conclusions. To achieve this, deep learning applications use a layered structure of algorithms called an artificial neural network. The design of an artificial neural network is inspired by the biological neural network of the human brain, leading to a process of learning that is far more capable than that of standard machine learning models.
It is difficult to guarantee that a deep learning model will never come to any incorrect conclusions. Making the models accurate requires imparting lots of training during the learning phase. The new deep learning models use artificial neural networks with many layers of processing units. These models are trained exhaustively with massive amounts of data. They take advantage of advances in computing power and improved training techniques to learn complex patterns in large amounts of data. When it works as intended, deep learning is considered to be a scientific marvel and the very backbone of AI.
Through his creations, Krishna C. Mukherjee has greatly enhanced the automation and prediction capabilities of AI. He and pioneers like him have brought about the Fourth Industrial Revolution, in which AI is redefining our lives.