Government officials emphasized testing and reliability and urged caution in introducing AI. Data preparation, structured formatting, and AI awareness are key.
Rohit Bhardwaj, deputy director, Department of Data Informatics and Innovation, Ministry of Statistics and Program Implementation (MoSPI), also suggested that government departments need to create computer-readable files to make relevant data AI-enabled.
“There should be context files, semantics, metadata,” he said, emphasizing storing information in a structured format that is AI-ready.
Bhardwaj pointed to the importance of testing AI solutions before use, citing a report by a group of researchers at a Canadian university showing that AI can analyze a given dataset in multiple ways, even when given similar prompts.
“I just want to warn you that you shouldn’t just make a fuss about something that hasn’t been tested yet,” he said during a session on “AI-enabled data: shared infrastructure for innovation” at the AI Impact Summit.
“I’ll be the first to implement AI in my work, but AI needs to be trustworthy,” he said, adding, “People don’t understand what it takes to make data AI-enabled. It’s the responsibility of institutions like MOSPI to make people aware of what AI-enabled means.”
He suggested that all flies should not be in PDF format as ministries and government departments should have catalogs and should be machine readable.
“I’m going to create a slide deck of what the AI can see and what the AI can’t see. So if I have 10 versions in my folder, some answers will come from version 1 and some answers will come from version 2. Unlike humans, AI is designed to scan the entirety of what’s available (data),” he said.
Google’s Prem Ramaswami said his company is looking to combine multiple data sets from around the world into a common knowledge graph and put a date search engine on top of it. “We’re open sourcing that entire stack so that you can access that data right away.”
The idea that data is centralized in one source is dangerous, he noted.
Instead, he suggested, data should be located in every organization, managed locally by the organization, and made accessible to businesses at an affordable price.
“When you have 74 million MSMEs in India, you cannot afford to hire data scientists or computer scientists,” he said.
He added: “If you’re a policymaker and you’re thinking about poverty, climate change, education, health, these are holistic issues. It’s no longer… you can go to one ministry and pull one spreadsheet and solve poverty. We should approach AI as a tool that we can use to come up with answers.”
