AI stands for artificial intelligence. It can also stand for advantage India. The Centre has planned to build a version of a foundational model for artificial intelligence (AI) which will be customised for use by Indian companies, entrepreneurs, academics and researchers, The Economic Times has stated.
Quoting officials who chose to remain anonymous, the report revealed that an amount of Rs 2,000 has been earmarked to launch the project which will take place after the general elections are over. Incidentally, the US and China are already building foundational models for the benefit of the general public. For the purpose IndiaAI Innovation Centre might be set up by the Ministry of Electronics and Information Technology under the Rs 10,000 crore IndiaAI Mission. Defence and agriculture are two specific areas that might immensely benefit from such an approach, the officials said.
“The government will likely tap eminent higher education institutes and prominent researchers working on AI in the private sector to work on foundational model,” a senior official said. This bureaucrat added that it might be a large action model (LAM) or large multimodal model (LMM). The architecture would be tailored to benefit a diverse range of services.
Foundational models are pre-trained generative transformers that are the starting point or fountainhead for the development of other AI models. These models analyse and breakdown old data in different ways to generate new response and models depending on how the user wants to utilise it.
“The needs and specific demands of India are very different from other companies globally. This (foundational model) will aim to provide output in more than one native language, borrowing from all the work that has been done so far on projects such as Bhashini,” the official told the newspaper. India being a land where people speak and write dozens of languages and have needs that are peculiar to it. For example, Bhashini is an AI-based language translation platform which was developed by the IT ministry in 2022.
The Indian government is expected to employ publicly available data, digitised records of books, journals and research papers from public domains libraries to develop this model. It would also utilise any non-personal data that would be voluntarily given by companies, startups or researchers. “There are very obvious privacy concerns as well as copyright issues that come with data (used to train foundational models). So historically accurate data from books that are peer-reviewed, scientific research journals can be utilised. We may also look at a platform exclusively for Indian startups where non-personal and anonymised data can be volunteered for training of the model,” a bureaucrat told the newspaper.
This huge amount of data is needed for processing so that the model developed can have sufficient cases to study before the foundational model is declared fit for commercial use. “Consider, for example, all the data generated through and during the delivery of government welfare schemes, both at central and state levels. With appropriate and targeted training, the same anonymised data could be repurposed for a better and more targeted delivery of these services and could result in a much faster turnaround time even for private companies,” said a bureaucrat.
Furthermore, the foundational model will also be trained on publicly available datasets overseas and open-source tools for machine learning. “No such foundational model can be made in isolation. While we aim to build and train this model to be India-specific, the neural networks behind the learning engine will need all the data it can consume,” another official told the newspaper.
The world over both governments and private companies build foundational models. Data from the Stanford Center for Research on Foundation Models indicate that till April this year, there were more than 330 foundational models. While some of these were developed by governments of different countries, other were create by private sector companies.
Private companies such as OpenAI which is supported by Microsoft, Amazon and Google, Amazon have developed advanced forms AI foundational models.