homebusiness Newscompanies NewsAmazon's 'Guardrails' can help Indian government protect responsible AI practices, says AWS' chief technologist

Amazon's 'Guardrails' can help Indian government protect responsible AI practices, says AWS' chief technologist

AWS made headlines a week ago, when it announced the launch of 'Guardrails for Bedrock', which ensured that enterprises could be deemed capable to define their own standards of safety, including setting hate filters in LLMs or Large Language Models. 

Profile image

By Jude Sannith  Dec 12, 2023 11:32:26 AM IST (Updated)

Listen to the Article(6 Minutes)
3 Min Read
Days after Amazon Web Services (AWS) CEO, Adam Selipsky, announced 'Guardrails for Bedrock' — a feature that helps filter output from AI models — the company's chief technologist in India, Shalini Kapoor, told CNBC-TV18 that the tech could play a key role in helping the government ensure Responsible AI.

Share Market Live

View All

"We work with the government and we always give our point of view in so far as furthering our use of technology going forward," said Shalini, who is Director and Chief Technologist (Public Sector) at AWS India.
"Now, it's all about Guardrails and working with the government in consultation, even as we look at the legal and policy part of it (Responsible AI)," she added, "AI innovation and development is important, but what is also important are safeguards while innovation and development takes place."
The AWS' Director and Chief Technologist was responding to a CNBC-TV18 question on whether the launch of Guardrails meant that tech companies like AWS would bat for a more self-regulatory approach when it came to ensuring Responsible AI. She emphasized that a consultative approach from the government, and a collaboration between government and industry was the way forward in so far as AI regulation is concerned.
AWS made headlines a week ago, when it announced the launch of 'Guardrails for Bedrock', which ensured that enterprises could be deemed capable to define their own standards of safety, including setting hate filters in LLMs or Large Language Models.
This, in turn, means that the choice of output from an AI model would lie in the hands AWS customers, many of whom are large enterprises. "AWS is clear that we do not intend to use this customer data to fine-tune our models," said Shalini.
Ever since Selipsky's announcements at AWS' annual 'ReInvent' conference in Las Vegas, the company is banking on its new AI tool, Amazon Q to boost developer productivity, and help companies devise customized solutions and applications. Shalini said the launch would find great relevance in the Indian public sector as well.
"India is going to be a one-trillion-dollar digital economy, and all these announcements, including Amazon Q, can help encode transformation, help with Generative AI, developer productivity and new solutions for the public sector," she said. "Our AI stack has been a journey for the last 17 to 20 years, today it is complete because we are talking about the processes and infrastructure to build Generative AI and LLMs, and use them."
AWS has been working with the Indian public sector to develop platforms like CoWin which helped administer 2.2 billion doses of COVID-19 vaccines to Indian citizens, as well as DigiLocker, which has thus far stored 60 terabytes of digitized documentation of 189 million Indian citizens. The company has also been focusing on sectors like agri-tech, state government, and education. Shalini said AWS would continue laying emphasis on national platforms, even as it renews its focus on healthcare.
"The way citizens use healthcare facilities is set to transform especially in line with the vision that the government has laid down," she said, "We want to work with the government in creating a unified health interface, and health IDs using Amazon Health Lake in India, which we launched last year."
For context: Amazon Health Lake allows patient data to be stored in one location for monitored and regulated access. "We have also been working with the Centre of Microbiology to reduce the time cycle to conduct gene mapping, bringing this number down from 550 days to just 9 days," Shalini added.

Most Read

Share Market Live

View All
Top GainersTop Losers
CurrencyCommodities
CurrencyPriceChange%Change