Mint DIS 2023 | Explainable AI is critical for success of AI ecosystem: Kashyap Kompella 

Explainable artificial intelligence (XAI) can make AI applications much better and also boost their adoption, said Kashyap Kompella, CEO of RPA2AI Research, at the Mint Digital Innovation Summit & Awards on June 9.

“Explainable AI is really the X- factor for the success of the AI ecosystem,” said Kompella. 

XAI is a set of processes that explains how AI models make decisions so people using them can understand how they work.  

Kompella noted that the widely used black box model only gives the answers but doesn't show the workings. 

He explains that the objective of XAI is to understand what's going on in the internal workings of neural networks so it’s understandable first to a professional who can then explain it to their customers or employees. 

“So far, all the efforts of the community have been focused on making AI explainable to the technologists. There are people who are actually going to implement these systems in their businesses. They need to be comfortable with whether they can adopt it or will there be any legal consequences if it doesn't work as advertised,” said Kompella. 

This will also make AI systems more transparent and make people trust them more. “Broadly, we want the AI systems that we're deploying in the world to be accountable and to be transparent,” said Kompella. 

Even though neural networks are software, Kompella said that when people use AI systems for automating decisions, they have an impact on the end users, which is why it’s important that they work as they're advertised and there are no unfair outcomes that can harm users. 

Kompella also reminded that XAI is important from a regulation point of view as there is a lot of regulation that is coming in this direction. 

“Niti Aayog has these responsible AI guidelines, which say that explainability is a key part of the systems that you are building.  GDPR has a right to explainability. From July 15 onwards, New York has an external audit requirement for AI systems and to be able to do these audits for HR AI systems, you need explainability as a key feature,” said Kompella. 

To make XAI part of every AI system, Kompella feels that it should be a part of the “entire lifecycle of an AI system” and should not be treated as one stage of it.  

He further said that the need for XAI presents a big opportunity for the Indian tech ecosystem. 

“The secret that nobody tells you about AI is that AI is automating certain things, but to get to that it's a huge service and consulting opportunity for explainable AI practice. The AI audits require a lot of explainable AI tools. So it's a huge white space for investors and startups alike,” he added.