Murari Sridharan spoke about Machine Learning (ML) and Artificial Intelligence (AI) in his lecture at ACJ-Bloomberg on August 29. The Chief Technology Officer of BankBazaar.com explained what it means to train a computer.

ML is a subset of AI, he said. Most companies and their models actually do not apply AI in its true form and just apply ML, he said. Tesla, Google, Microsoft and Facebook are the only companies that are actually using AI, he said. Most people are actually talking about ML, when they claim to be using AI.

Sridharan ran the team that launched Microsoft’s digital assistant Cortana. He said that the User Experience (UX) research team for the project spoke to the personal assistants of people holding high posts in several organizations to develop Cortana.

The recent transformation of AI and ML can be equated to that from Roman to Arabic numerals. All ML systems either do production or classification, he said. Human intelligence will still be necessary to identify errors.

Sridharan used three-dimensional scatter diagrams to illustrate the basic operation of ML. Using the example of an apple, he said that when we teach a machine that an apple is red, it can make the mistake of classifying all red things as apples. When we add further classifications, we can teach the machine to distinguish based on more specific characteristics. He illustrated this by adding more dimensions to a graph.

He said that we need something that truly learns wherein we do not have to change the rules manually. In the current AI and ML setups we have, human intervention will always be necessary. For such systems, every input needs to be converted into data. Large language models are used to train data. Humans, he said, will always have a better level of understanding and the most AI can do is complete tasks of humans as per our instructions.">