Earlier today, Google announced a grant of 1 million USD to IIT Madras for setting up a centre for Responsible AI. The initiative will look to mitigate bias using Natural Language Processing (NLP) models and enable fair use of AI, particularly in the Indian context.
The partnership was announced at the Google for India 2022 event, which witnessed the declaration of many such programmes in their commitment to building a more inclusive, helpful, and safer internet for every Indian. In addition, responsible AI was also a highly focused theme in Google’s first AI@ event in November. The theme thus continues to be a highlight in the India event as well. The models will be open-sourced to collaborate with the industry for faster progress.
Prof. Ravindran, Center for Responsible AI, IIT Madras, said, “There is a need to develop technology that is understandable to the end users, providing strong safety and performance guarantees, and to raise awareness of the impact of the technology on society.”
Additionally, Google is also working on building AI-led security solutions to respond to the complex cybersecurity challenges we face today. It is working with the National e-Governance Division and Ministry of Electronics and Information Technology (MEITY) to add Digilocker-verified documents straight to the Files by Google App for Android users for quick and safe access to the documents.
As part of building security features, it also introduced enhanced security features on Google Pay, including multi-layered intelligent warnings to alert users if the fraud detection system detects suspicious activity.
Furthermore, Google will also collaborate with the Government of Telangana to bring AI-based solutions for agricultural innovation. Google’s partnership with Croprin, a global ag-ecosystem intelligence provider, will lead this to bring about applications that can track land use and bring digital services to farmers nationwide.
Finally, the event also introduced ‘Project Vaani’, a joint initiative by Google and the Indian Institute of Science (IISc), which will collect and transcribe open source speech data from all 773 districts of India and makes it accessible through the Government of India’s Bhashini project.
In the 2022 AI@ event, Google mentioned that it plans to build a model that supports 1000 most-spoken languages through its 1000 Languages Initiative. A glimpse of that was also seen here, where it has made the Search Results Pages bilingual by tapping into their advanced ML-based translation models and cross-language search technology. In addition, a set of AI-based product integrations are enabled, which will have a new, evolved, multimodal, bilingual model covering over 100 Indian languages.