Artificial intelligence (AI) models should be designed and trained to address particular problems, or for specific use cases, Infosys co-founder Nandan Nilekani said, while emphasizing on developing AI systems that can help make a difference in people’s lives.
Nilekani, who built Aadhaar as the chairman of the Unique Identification Authority of India (UIDAI), said AI can be used in digital public infrastructure (DPI) if it is a small model trained on relevant data.
Speaking at the Carnegie Global Technology Summit 2023, Nilekani said, “AI strategy should be use-case-led. They should be ‘outside in and not inside out’. Most of the AI conversations happening today are ‘inside out’, which says ‘my model is bigger than yours’.”
“All this is great but how do we make a difference to the lives of people? How can the common man improve his life with AI? We can only do that by having very good compelling use cases that make a difference to people’s lives, which is the whole DPI philosophy,” Nilekani, who is currently the non-executive chairman of Infosys, said.
A small model trained on relevant data gives you better results than a generalized model which does everything, he said.
Talking about India’s capability of using AI for public good, Nilekani said, “Digital public intelligence with private innovation will make India the AI use-case capital of the world. We are not in the ‘we will build a bigger model than you’ race. That’s of no interest to us. We are in the business of using this stuff to make a difference to a billion people.”
The Aadhaar pioneer said AI models can even be trained for various languages, which can then be used by different service providers through mobile applications.
“If we can fix the language and if the language is available to you as open data, then you and your fintech application can use that language to provide credit or whatever you can do to that person in that language including in a mixed language. The AI has to understand our unique behavior,” he said.
“That will amplify your ability to deliver your products to more consumers. Think of it as an infrastructure at scale, low cost, super efficient, low-inference cost, fully open source which can plug into your apps,” Nilekani added.
Loading the player...
Understanding the climate crisis with Marina Romanello
More Top Stories:
US regulator proposes new rules to combat AI-generated deepfakes