<p>The digital revolution has changed the world significantly in the last decade. Algorithm and computing capabilities are the ingredients behind the massive digital revolution. The complex algorithms have helped in building critical applications and software to ease our day to day activities. Algorithm along with massive computing infrastructure helps in solving critical problems in real-time, and help in forecasting and prediction. We are in Second Machine Age (according to the authors Erik Brynjolfsson and Andrew McAfee) where the machines can think and auto-correct themselves. Even after decades of research and study, the field of Artificial Intelligence AI is still one of the elusive subjects in computer science for its massive scope ranging from medicine and space travel to warfare. Enterprises are turning to artificial intelligence (AI) to help analyse data and use it to make effective business decisions and solve business problems.<br><br>The term artificial intelligence was first coined by John McCarthy in 1956. It has come a long way from science fiction movies to become a reality today. The machines are no more just smart - they are becoming super intelligent with the ability to think and react like humans. The AI can be divided into 3 major categories:<br><br>" Artificial Narrow Intelligence (ANI): Mostly referred to as weak AI. Artificial Narrow Intelligence is the AI that specializes in a single area of operation. e.g. a calculator program can work just to calculate numbers. It cannot help in setting a reminder for something.<br>" Artificial General Intelligence (AGI): AGI is referred to as strong AI. This AI can be as strong as a human brain. The machine can perform any intellectual work that human being can. Building an AGI is much tougher than building an ANI. e.g. IBM Watson, Wipro Holmes<br>" Artificial Super-intelligence (ASI): ASI is an intellect that is much smarter than humans. It can range from computers which are little smarter than humans to computers that are trillion times more smart.<br>There are many factors driving the growth in machine learning:<br>" The volumes of data generated by social media, digitization of financial services, rise of e-commerce, increased use of M2M devices and sensors and data generated from connected devices.<br>" The complexities of connected subsystems and their interactions result in system dynamics that can no longer be fully comprehended, even by the smartest engineers.<br>" Traditional system engineering has become a bottle-neck in delivering cost-effective solutions.<br>" The availability of less expensive in-memory storage, faster computer hardware and easy-to-use cloud solutions.<br>Current Scenario<br>As per a study by BCC research, the global market for smart machines will grow to $15.3 billion by 2019 with an average annual growth rate of 19.7%. The smart machine market is divided into 5 segments:<br>" Expert Systems: Help users solve complex issues by providing recommendation derived from their own knowledge base. e.g. medical decision support systems, smart grids<br>" Autonomous Robots: They do not need human assistance and move freely. They have different degrees of autonomy. e.g. spirit rover, Opportunity rover, Topio (humanoid robot) <br>" Intelligent Assistants: They make their own decisions depending upon the issues. e.g Apple's Siri<br>" Embedded Systems: They are present in different machines where they control their function and make useful applications possible.<br>" Neurocomputers: They work as human brains. The neurons trigger one another and learn to recognize patterns and interrelationships.<br><br>Wide range of research activity is undergoing to make machines super intelligent so that they can take their own decisions based upon historical data, used cases and other resources. Machines are slowly entering different sectors such as medicine, manufacturing, national security, etc. Amazon, Google, IBM and Microsoft are the biggest players trying to dominate the machine learning market. IBM Watson is a cognitive computing system which understands natural language and is built to mirror the same learning process like the humans i.e. Observe, Interpret, Evaluate and Decide. Similarly, Wipro Technology has developed a cognitive computing system called Holmes - a cognitive machine learning platform to help a company improve its internal efficiency and productivity. TCS is also working on machine vision and video analytics to improve Retail Market Operations and Bank Automation. These intelligent video analytics can be used to understand the customer behaviour and offer actionable insights.<br><br>Other significant advancements in machine learning are Google's Prediction and Microsoft's Azure machine learning. Voice recognition and personal assistance systems like Apple's Siri and Microsoft's Cortana and Google's Now are gaining popularity in the mobile platform.<br><br><strong><span style="color:#800000;">Used cases of Artificial Intelligence</span></strong><br>AI has become a reality and has wide scope of use even in its early stages of development. The super intelligent machines are making complex decisions in different functional areas across the industries.<br><br>" Health Care: The computer-aided diagnostics are gaining popularity in the healthcare sector. The intelligent machines can take patient's condition such as vital signs, symptoms, lab tests or toxic exposure to provide disease classification or even recommended therapy.<br>" Environment: Smart machines analyse sensor data in hazardous environments such as measurement of air quality, equipment performance or employee productivity, or even some typical behaviour to predict the likelihood of accidents. This application has been widely adopted to alert truck drivers for potential accidents. Stanford university researchers are using machine learning AI program along with NASA's collection of solar observations to predict solar flares of the Sun. The ORCHID project, a consortium of UK universities and private firms aims to streamline disaster response by combining human and artificial intelligence into an efficient complementary unit known as a Human Agent Collective (HAC). The computer systems being developed can assume tasks such as directing surveillance drones, resource management and search planning.<br>" Public Services: Machines can estimate criminal activities like public offence, death, murder and theft by taking inputs of the current condition of a city. i.e. weather, season, traffic, sport events, recent crimes and prisoner releases. It can help in improving the security resources e.g. police.<br>" Smart Cities and Smart Transportation: Machines can help in traffic optimization by analysing traffic patterns using sensor data, accidents and roadworks. They can predict delays or road obstructions and can recommend faster route for public and commercial vehicles.<br>" Retail: The machine model can help in recommending product to customers. The propensity-to-buy model analyses customer profile i.e. customer information, activities, recent purchases, etc. to understand the buying behaviour and recommends products. It also determines the churn rate. Machines can analyse the historical sale data, competition, weather and customer profile to estimate demand for a product and thus the price.<br><br><span style="color:#800000;"><strong>Limitations</strong></span><br>Though machine learning and AI are helping humans in everyday activity and they possess a wide scope of usage, they have some limitations too.<br><br><strong>Security: </strong>Security is a concern as there are threats of hackers and spammers who can alternate the controls of an intelligent machine to create hostile situation.<br><br><strong>Lack of regulations: </strong>The AI and machine industry still do not have proper regulation and policies. There needs to be strong policies and regulations around full automation in different industries.<br><br><strong>Healthcare diagnosis:</strong> A proper diagnosis of a patient needs analysis of structure, unstructured data, image data (X-Ray) and visual cues from the patient. At the early stage, the AI machines cannot fully analyse the situation.<br><br><strong>Coding error: </strong>A simple coding error may result in unexpected outputs sometimes which minimizes the usability of AI in sensitive areas such as national security, healthcare diagnosis, etc.<br><br><em>The author, Hemant Joshi, is a partner with Deloitte Haskins & Sells LLP</em></p>