In line with India’s AI Mission, who will have access to GPUs?
Speaking on the artificial intelligence (AI) project that was approved in March of this year, Singh stated that researchers and companies developing answers for the government-floated problem statement using Indian datasets will all have access to the computing. “We will prioritize the problem statements that we have identified, such as the health sector, education, agriculture, and social impact, but startups may be able to come up with other solutions that we may not be able to,” he stated. He also said that there would be enough computer power to meet all of the nation’s requirements.
Following the conclusion of the model code of conduct (MCC), according to Singh, the request for proposals (RFP) for Graphics Processing Units (GPUs) would be made public. This request for proposals is for the establishment of a supercomputer infrastructure with all the chips that the industry needs. The computing power supplied will depend on the needs of the users, he explained. “I won’t say it will only be NVIDIA, it will only be H100s, it can be AMD 300, it can be Gaudi 3, it can be a combination of the three,” he stated. He said that in an effort to ascertain the requirements of the people, the administration is surveying users.
It will take some time for such infrastructure to be provided, even when the RFP is issued. He stated, “It will take anywhere from 12 to 18 months to become actually operational.” Other players are putting up AI computing infrastructure till this infrastructure is in place, and a system will be in place to enable those who require said infrastructure to obtain it at a discounted price.
Using Artificial Intelligence (AI) in Digital Public Infrastructure (DPI): Singh stated that the type of data produced by DPI apps presents a significant opportunity for improving solutions through the addition of an AI layer. For instance, the agriculture ministry generates a lot of data for various purposes such as crop research, Kisan call center, and so on. They also use PM, Kisan, and DBT [direct benefit transfer] for money transfers. He clarified, “One may create applications on top of that using that data to make sure farmers can receive alerts on voice in their home tongue.
Because they feel uncomfortable downloading apps or surfing websites, the majority of people in India are unable to use DPI applications. But according to Singh, it would be revolutionary if the same individuals were offered voice-enabled applications via natural language processing (NLP).
How can deepfakes be dealt with by the government?
According to Singh, there is a global agreement to develop technology that are capable of identifying deepfakes. “Can we have provisions where something, if an image or video is generated by AI, then to mention that this is AI generated? Can we also have provisions where things are labeled, watermarked, and deepfakes so the people should be able to know?” Singh asked. Naturally, he said, it would be difficult to ensure compliance with all of this, and in the end, the law would need to be involved.
Singh stated that as rules prohibiting deception have existed since the invention of Photoshop, they may also be applied to artificial intelligence (AI) through improved training for law enforcement and police personnel. He said, “It would not mean much if we just made regulations and didn’t have the kind capacity or the means to enforce them.”
Will the translation feature of OpenAI affect Bhashini?
According to Singh, it’s a positive thing that OpenAI’s translation capability now supports Indian languages. This is what’s going to occur. And this will occur as time goes on, computing power increases, data sets grow, and model power increases, leading to far improved models. Therefore, Bashini will also need to use the various instruments that are available to them to step up their game,” he said.
