Understanding the buzzwords of Artificial Intelligence & Machine Learning

As the Co-founder and CTO of a tech-driven real estate start-up, I am frequently asked a question: how much of an impact has the technology, Artificial Intelligence & Machine Learning, made on the way businesses operate today?

Rather than answering this with mundane terminologies like disruption, innovation, automation etc., I would shift focus to the impact technology has brought in our day to day lives. When a local paan wala accepts money through UPI apps, you cannot be wondering whether we need technology or not to run a business. Technology has always transformed the way business operates, in the form of Machinery removing mechanical labour in the past to AI systems removing Human intelligence today.

No technology is getting as much share of the airwaves as Artificial Intelligence (AI) and Machine Learning (ML)– two of the most era-defining developments to have emerged in recent years. But, despite the frequency of discussion, many professionals don’t fully understand these technologies. Some use AI and ML interchangeably as synonymous terms, while several others treat them as separate technologies altogether.

The truth, however, is somewhere in between. AI and ML are indeed related, but not how most people think they are.

Back to the basics: An introduction to Artificial Intelligence and Machine Learning

Providing an exact definition is challenging because the AI goalpost is constantly shifting. In 1997, IBM’s Deep Blue – a computer that defeated the reigning chess champion Garry Kasparov under standard tournament rules – was considered to be the epitome of AI. Today, we’re talking humanoid robots, virtual machines & self-driving cars.

The general consensus amongst domain experts is to describe AI as a field of Computer Science dealing with incorporating human intelligence into machines. It is the ability of a machine to display cognitive capabilities of a human mind such as learning, problem-solving, creative thinking etc.

1. These can be analytical systems that make predictions based on observations from the past events. For Example: Using historical trends of rental prices to predict the rental prices of new houses.

2. These can be human-like systems that combine previous experiences with emotional intelligence while decision-making. For Example: Chat Bots that can let you resolve a grievance with a service.

3. In the future, these might be self-aware systems that are capable of combining cognitive, emotional and social intelligence while decision-making. For example; let’s say in the future we may be able to replace an entire customer support executive with an AI agent.

Machine Learning is a branch or a subset of AI. ML is just a way by which we approach some of the solutions in AI. ML algorithms derive the statistical representation underlying the data without being explicitly programmed about it. This machine will then be able to make inferences for new sets of data.

Classical ML Vs Deep Learning

In classical ML, we have a prediction to make about a target variable, and we handpick a set of independent variables(features) that most likely influence the target variable. We then ask the machine to learn the relationship between the independent variables and the target variable. For example, an ML algorithm can learn the relationship between the rent of a house with its attributes like locality, size, age etc.

Recent breakthroughs in the field of Machine Learning has enabled us to learn patterns represented by data without handpicking features. These class of machine learning techniques are called Deep Learning.

Deep Learning enables us to learn abstract representations in data. This is required to understand data in the form of images, videos, audios, text etc. Otherwise, it is difficult to construct generalized rules that can understand patterns in such form of data. For example, we can never write a well-describing logic that identifies animals in an image. This would involve permutations & combinations over thousands of pixels. Deep Learning enables to learn abstract features in the data that is necessary to answer what is dog and cat in a picture of animals

Impact of ML/AI

While some businesses have been extremely successful in bringing value to its users with ML, many others have failed miserably. Businesses fail to derive value out of ML when they try to do ML for the sake of ML. The key to success here is identifying the exact problems that machines can solve for you. Machines are capable of doing things faster, better and more efficient. Image intelligence systems remove manual overhead of reviewing real estate images at NoBroker. NLP and ASR systems are now capable of addressing customer grievances better. The applications are too many.

To summarize, AI is a broad spectrum of practices dealing with Machine Intelligence. And ML is one of the approaches using which we solve some of the problems defined under AI. And Deep Learning is one of the approaches to realizing ML.

This clarification is necessary as many professionals without hands-on expertise in the subject seem to get confounded with these buzzwords. AI has become the most popular technology of today. ML is solving some of the problems in AI today in a specific way. Tomorrow we might see AI solutions dealt in a way different from how it is dealt with today. As Judea Pearl - a pioneer in the field of AI recently pointed out, AI has been stuck in a decades-long rut; and the way to progress is to teach the Machine to understand the question “Why?” - Machines being truly able to answer questions of intelligence & creativity. This may be far, but the process is already in going and the impact we shall wait and watch.

By Akhil Gupta, CTO and Co-founder, NoBroker.com