By Navaneet Mishra, Senior Vice President and Head, Hexagon R&D India
The visibility and hype of Generative AI (GenAI) has certainly taken over our collective imagination, casting the applications of classical Artificial Intelligence (AI) into the shadows. Not surprising because, as humans, we are naturally fascinated by human-like responses to queries and a technology’s ability to write essays and generate images. But classical AI such as Machine Learning (ML) has been playing a huge role in solving core, critical, and complex real-world problems. Many of these cannot be handled by humans, or at least not meaningfully.
Understanding the difference
GenAI is a branch of AI that fetches information and stitches it neatly within the context of its use, going beyond the pattern recognition and decision-making that we expect from classical AI. GenAI’s value lies in being able to predict the next piece of data – whether word, pixel, or musical byte – based on the last such data (token).
For example, in text generation, based on its learning, GenAI knows the statistical likelihood of the next word following the current word. The best GenAIs deliver “delightful” results because they are trained on vast and varied amounts of data in the magnitude of billions – hence the moniker Large Language Model (LLM).
Machine Learning, on the other hand, implies two things: (a) By measuring various characteristics (dimensions) of objects, we can write a program to closely identify / categorize the object; (b) The more volume and diversified data we have initially, the greater accuracy we will achieve.
For example, if we have a basket of fruits, and the objective is to identify the fruit, then we can create a program that consumes measurements of length, width, weight, color, texture, and so on, and in turn tells us what fruit it is or it could likely be. In its simplest form, this is how ML works.
The real applications of it are way more complex and find use in areas like urban planning, public safety, industrial automation, and maintenance of critical infrastructure, amongst others.
According to a report by MarketsandMarkets, the global ML market size is projected to reach $96.7 billion by 2025, growing at a CAGR of 43.8% from 2020 to 2025.
Let’s explore some of the key applications of ML in advancing our world today.
Classifying satellite data for informed urban planning
With more than 800 million people expected to move to cities in the next 25 years as per UN studies, urban planning, with the right green coverage, is a big global challenge. But with satellite imagery, today, we have airborne sensors that collect data and capture images of specific cities. If we want to know the green coverage and the infrastructure distribution of a city from that huge data set, it is a classic use of ML.
The hundreds of gigabytes of laser imagery that the sensors would have collected can be recognized by AI algos as buildings, tree covers, streets, water bodies, large infrastructure, and so on. This recognition, categorization, and mapping is critical for urban planning and is the only way that civic bodies can make informed decisions and simulate possible solutions to arrive at the best one.
Video analytics for early threat detection
Around the world, threats to public safety are rising. As per Grand View Research, during 2022 to 2027, the public safety and security market is set to grow from $433.6 billion to $707.2 billion, at a CAGR of 10%+. From a solution point of view, it would be smart and useful if we detect threats as they are unfolding.
Video analytics come in handy for that. These AI models can recognize specific individuals (face detection and recognition), understand emotions from facial expressions (emotion detection), identify objects (object detection), and actions that can lead to criminal activities (anomaly detection), and so on. These can also do crowd estimation.
All of these need deep ML technologies. Direct human action or even conventional programming cannot address these situations at speed or scale.
Predictive maintenance for industrial efficiency
Industrial plants are incredibly capital intensive, and there need to be preventative measures to make sure that the plants continue to operate. AI and ML are critical technologies that help maintain productivity with predictive maintenance capabilities.
Sensors and IoT devices are deployed to capture vibration, pressure, air quality, temperatures, and such parameters. Comparisons between current and historical data then throw up patterns and correlations that are used to predict failures – and thereby avert them.
Today’s most advanced set-ups use data-rich digital twins of heavy machinery enabling the maintenance teams to run simulations with real-time data and gather way more information about risks, concerns, and potential failures than was possible ever before.
Looking beyond the hype
GenAI is cool. It aids productivity and can make or save money and time for the entertainment, design, and content creation industries. It is however not the only nor the only important application of AI. Classical AI such as ML and Deep Learning will play key roles in creating huge advances in many aspects of our lives. We need to continue to invest in these technologies to tackle pressing and emerging social, industrial, and governance issues effectively. These may not be as in-your-face as GenAI but they will be critical cores of innovation to move the human race forward.
