The government plans to develop its own version of the basic artificial intelligence (AI) model. According to an Economic Times report, the proposed core AI model will be adapted for exploit by Indian companies, entrepreneurs, scientists and researchers.
Citing people aware of the matter, the report states that the union government has allocated Rs 2,000 crore for the ambitious project, which is likely to be launched after the ongoing general elections.
Sources said India’s core model may be led by the IndiaAI Innovation Center, which will be set up by the Ministry of Electronics and Information Technology under the Rs 10,000-crore IndiaAI Mission.
“The government is likely to recruit eminent higher education institutions and eminent AI researchers in the private sector to work on the core model,” a senior official said. This could be a enormous action model (LAM) or a enormous multimodal model (LMM), so that the results can be used in a wide range of applications and services, he added.
What are base models?
Core models, as explained on the Amazon Web Services website, are a form of generative artificial intelligence (generative AI). They generate output from one or more inputs (hints) in the form of instructions in human language.
“India’s needs and specific requirements are very different from those of other companies around the world. “This (basic model) will aim to deliver results in more than one native language, borrowing all the work done so far in projects like Bhashini,” a senior official said.
Both private companies and public governments can develop the basic model. To date, private companies and governments have developed more than 330 foundation models, according to the Stanford Center for Research on Foundation Model.
The government will exploit public data to train the base model
The report stated that the Government of India plans to exploit publicly available data, digitized records of books, journals and research articles from public libraries to train the model. It will also exploit any other anonymized, non-personal data voluntarily provided by companies, startups or researchers.
“There are very obvious privacy concerns as well as copyright concerns related to the data (used to train the underlying models). You can therefore exploit historically right data from peer-reviewed books and scientific journals. We may also consider a platform exclusively for Indian startups where non-personal and anonymous data can be voluntarily shared for model training,” an official was quoted as saying in the ET report.
Additionally, the base model will also be trained on global, publicly available datasets and open source machine learning tools.
Citing people aware of the matter, the report states that the union government has allocated Rs 2,000 crore for the ambitious project, which is likely to be launched after the ongoing general elections.
Sources said India’s core model may be led by the IndiaAI Innovation Center, which will be set up by the Ministry of Electronics and Information Technology under the Rs 10,000-crore IndiaAI Mission.
“The government is likely to recruit eminent higher education institutions and eminent AI researchers in the private sector to work on the core model,” a senior official said. This could be a enormous action model (LAM) or a enormous multimodal model (LMM), so that the results can be used in a wide range of applications and services, he added.
What are base models?
Core models, as explained on the Amazon Web Services website, are a form of generative artificial intelligence (generative AI). They generate output from one or more inputs (hints) in the form of instructions in human language.
“India’s needs and specific requirements are very different from those of other companies around the world. “This (basic model) will aim to deliver results in more than one native language, borrowing all the work done so far in projects like Bhashini,” a senior official said.
Both private companies and public governments can develop the basic model. To date, private companies and governments have developed more than 330 foundation models, according to the Stanford Center for Research on Foundation Model.
The government will exploit public data to train the base model
The report stated that the Government of India plans to exploit publicly available data, digitized records of books, journals and research articles from public libraries to train the model. It will also exploit any other anonymized, non-personal data voluntarily provided by companies, startups or researchers.
“There are very obvious privacy concerns as well as copyright concerns related to the data (used to train the underlying models). You can therefore exploit historically right data from peer-reviewed books and scientific journals. We may also consider a platform exclusively for Indian startups where non-personal and anonymous data can be voluntarily shared for model training,” an official was quoted as saying in the ET report.
Additionally, the base model will also be trained on global, publicly available datasets and open source machine learning tools.
