Adding innovations at every turn, artificial intelligence, machine learning, and the Internet of Things are transforming industries. With strong government backing and academic potential, how will these technologies shape India’s future?
These days, almost everyone is excited to know about artificial intelligence (AI) and machine learning (ML). In the year 2020, nearly 20 billion objects have become a part of this network, and experts say that by 2025, it will be almost 80 billion. The applications of AI and ML have led to the development of remarkable systems that are now widely used across various industries worldwide, from agriculture and irrigation systems to weather forecasting, smart homes, manufacturing, e-commerce, and the automotive sector.
In India, the AI and ML market is projected to surpass US$15 billion, positioning the country as a significant player globally. This growth is being bolstered by the increasing number of startups and IT firms adopting AI, with a particular emphasis on ML and deep learning (DL) as core technologies for their products and innovations.
Why IoT?
The Internet of Things (IoT) links the physical and virtual worlds through objects equipped with sensors that connect to the internet via local interfaces like Wi-Fi. These objects share sensor data over a secure cloud platform, where computation and analysis occur. IoT relies on various technical components, each with specific capabilities. IoT objects include user interfaces that enable direct or indirect communication with users, often via smartphones or smartwatches, ensuring seamless human-technology interaction.
One such capability of IoT is communication and mutual interaction. Objects can network with internet resources or each other to access data and services and update their state. Wireless technologies such as GSM, UMTS, Wi-Fi, Bluetooth, ZigBee, and various wireless personal area networks (WPAN) protocols are crucial for enabling this functionality.
Addressability is another capability. IoT objects can be located and addressed through discovery, look-up, or named services, allowing remote interrogation or configuration without physical interaction.
Identification is critical as well. Objects are uniquely identifiable, linking them to specific information. Technologies like radio frequency identification (RFID), near-field communication (NFC), and bar codes enable the identification of even passive objects without energy resources. The data can be retrieved via a mediator like an RFID reader or mobile phone connected to the network.
Sensing capabilities are essential for IoT objects. Using sensors, they collect environmental data, which they can record, forward, or react to, allowing adaptation to surroundings. Actuation, on the other hand, enables IoT objects to manipulate their environment through actuators that convert electrical signals into mechanical movement, enabling remote control of real-world processes.
Embedded information processing allows smart objects to process and interpret sensor data using processors or microcontrollers with storage. They can store usage data for future reference. Localisation is another important feature. Smart objects can determine their physical location using global positioning system (GPS), and ultra-wideband (UWB), ultrasound, radio beacons, or optical technologies, aiding precise positioning.
All about ML
ML branches out from a science that deals with the development and study of algorithms that can think or perform tasks like the human brain, known as AI. ML comes with a set of tools to train an algorithm which can then be embedded into a machine or computer system to do some smart, intelligent analysis or computations.
It can be classified into three main types: supervised learning, unsupervised learning, and reinforcement learning.
Supervised learning involves training a model using labelled data. During the training process, the machine is fed data with known outcomes, allowing it to make predictions or classifications about unknown variables based on past examples. This technique is widely used in business applications such as customer segmentation, churn prediction, and recommendation systems.
Unsupervised learning, on the other hand, works with unlabeled data. In this method, data points are grouped into clusters based on similarities or behaviours, and analyses or predictions are made based on these clusters. It is commonly applied in fields like anomaly detection, clustering visualisations, and discovering associations between data elements.
Reinforcement learning is another method where a model learns by interacting with its environment and receiving feedback in the form of rewards. The model aims to take actions that maximise its cumulative reward over time. This approach is used in applications such as robotics for motion control, industrial automation, business strategy optimisation, and even medical devices.
DL, a subset of ML, mimics how the human brain operates by using a network of multiple layers of neurons. This multi-layered neural network is employed to train models similarly. DL can be applied using supervised, unsupervised, or reinforcement techniques and is based on architectures like convolutional neural networks, artificial neural networks, and recurrent neural networks. It has succeeded in applications such as computer vision, speech recognition, and language processing.
The multitude of applications across industries and society
Industrial applications for cost savings are increasingly relying on ML to enhance efficiency. Industries use ultrasonic sensors around their machines, generating millions of data points daily. These data are fed into ML algorithms, which learn what is necessary for the machinery to operate at its best. For example, Goldcorp, a mining company, operates numerous vehicles and machines to transport materials. A sudden breakdown of these machines can cost the company around two million dollars per day. By leveraging ML, Goldcorp can monitor its machines 24/7, minimising downtime and maximising productivity.
Smart home technologies have been evolving for over a decade, with manufacturers working on adding intelligence and connectivity to household objects and appliances. Electrolux, for instance, proposed the idea of an internet-connected fridge in 2000. Early innovators envisioned a world where consumers could remotely control their homes, checking whether they have enough milk, turning off the central heating, unlocking doors, and even checking on elderly relatives from virtually anywhere. Now, with the declining cost of wireless hardware, the expansion of mobile networks, new business models, and the widespread use of smartphones, this vision is becoming a reality.
Connected security systems are increasingly being used to enhance home safety. These systems typically incorporate sensors to detect when doors or windows are opened or when movement is detected within the property. Many of these systems also offer home automation features, enabling users to lock or unlock doors remotely via smartphones. Today, these systems extend to advanced security features such as face recognition to detect if a stranger tries to enter a home, smoke detection for fire or gas leakages and sound analysis. Companies like Rapid7, SimpliSafe, Ring, and Fluidmesh Networks are market leaders in applying ML to security systems.
In healthcare, devices and intelligent medical sensors are transforming how healthcare services are accessed globally. Wearable devices, widely used today, incorporate smart sensors that regularly monitor an individual’s health, including parameters like pulse rate. ML and DL algorithms analyse this data to provide intelligent insights based on an individual’s behaviour. This not only accelerates diagnosis and improves efficiency but also helps reduce the costs associated with doctor visits and medications.
Fig. 1 illustrates the three-tier architecture of a basic Internet of Medical Things (IoMT) system. The first level includes user interfaces in the form of wearable hardware, such as smartwatches and smartphones. These devices perform preliminary data processing and some level of analysis on the data collected from their body sensors.
The second level comprises the gateway, which may be field sensor networks or local servers/gateways. This layer is a middleware, facilitating communication between the first and third levels.
The third level represents the cloud services layer, where high-performance computing, data storage, and advanced analytics powered by ML and DL algorithms are used to generate valuable insights.
Similarly, ML/DL systems are used for various medical appliances like X-ray machines, MRI scanners, and CT scanners for diagnosing some kind of dysfunction in the human body without much manual intervention from a doctor. And these are just a handful of use cases among the virtually numberless possibilities that have been and can be unlocked using multidisciplinary aspects of AI, for fulfilling several necessary social and economic goals.
Technological challenges
While the possible applications and scenarios outlined above may be very interesting, the demands on the underlying technology are substantial. Progressing from the vast network of computers to the remote and somewhat fuzzy goal of an IoT is something that must, therefore, be done one step at a time. In addition to the expectation that the technology must be available at a low cost if a large number of objects are actually to be equipped, we are also faced with many other challenges, such as scalability, arrival and operation, interoperability, discovery, software complexity, data volumes, data interpretation, security and personal privacy, fault power supply and wireless communications.
Government of India initiatives
There are several government undertakings that encourage research and development (R&D) AI, DL, ML, and physical systems. One such is the National Mission on Interdisciplinary Cyber-Physical Systems (NM-ICPS). It will be implemented through a network consisting of 15 technology innovation hubs (TIHs), six sectoral application hubs (SAHs), and four technology translation research parks (TTRPs). The preliminary phase of the mission has focused on establishing hubs in key areas such as technologies for the Internet of Things and Everything (IoE), sensors, activators, and control, as well as databanks, data services, and data analytics. It also includes hubs in advanced communication systems, robotics and autonomous systems, and cyber security, mainly focusing on cyber security for physical infrastructures.
Additionally, the mission eyes several objectives, including knowledge generation through basic and applied research, technology and product development, and commercialisation. It also aims to meet industry requirements, foster international collaborations, and facilitate the transformation of technology from the lab to real-world applications. Additionally, the mission focuses on human resource development (HRD) and skill development in ICT and automation.
Besides the national mission, NIELIT Calicut, under the Ministry of Electronics and Information Technology (MeitY), has launched a certificate course on cyber-physical systems. NITI Aayog has introduced a national policy on AI titled ‘AI for ALL’. Furthermore, the India-Sweden collaborative industrial research and development programme on AI will be implemented by the Global Innovation & Technology Alliance (GITA) from India and Vinnova in Sweden. The Department of Biotechnology (DBT) has also announced a call for AI applications under the ‘Affordable and Accessible Healthcare – Big Data and Genomics’ initiative, which is expected to attract scientists and researchers from multidisciplinary domains, including data science, computational biology, statistics, ML, and DL.
Academic support
Educational institutions and 40 to 70 universities across India have started degree (BTech/MTech/PhD) courses and are offering short courses (certificate/diploma/post-graduate diploma) on AI and ML. The initiatives taken by universities and colleges will enhance the specialised manpower to compensate for the demand in the coming days. The awareness and outreach programmes funded by government agencies are executed by various establishments.
Looking ahead
IoT, AI, and ML are transforming industries and shaping the future. The gradual rise of the Industrial Revolution 4.0 will fuel the development of cyber-social, cyber-biological, and cyber-corporate systems, besides cyber-physical. These advancements will drive interdisciplinary innovation as the progress continues. Moreover, national initiatives like ‘Digital India’, ‘Startup India’, ‘Make-in-India’, ‘Skill India’, and Atmanirbhar Bharat are laying the foundation for technological growth and self-reliance, which will nurture a thriving ecosystem and a self-sustaining nation.
Author(s): Subhranshu Sekhar Samal is the CEO of Atal Incubation Centre-AUDF, sponsored by AIM-NITI Aayog, in New Delhi. Ashank Bharati is the R&D Lead (SDE-2) at Nuclei, based in Bengaluru. Amrita Samal is the Deputy Chief of IT, AI Airport Services Limited (AIASL), in New Delhi.
