Saturday, December 16, 2023

Next: ChatGPT from Ola, "Made


 BENGALURU: Bhavish Aggarwal, the founder of Ola, made the announcement on Friday of a "Made for India" ChatGPT alternative that, according to him, will support ten Indian languages. Similar initiatives are being pursued by a number of businesses in India, with some of them drawing inspiration from the government's Bhashini mission, which aims to create an Indian language database that can be used to train AI models.

Aggarwal's AI company, which will be known as Krutrim, which is the Sanskrit word for artificial, will have a distinct identity from Ola's EV and mobility businesses. Aggarwal said the organization was consolidated as Krutrim SI Plans in April with groups situated in Bengaluru and San Francisco.

In January, Krutrim will release a beta version that developers and businesses can use to create chatbots tailored to their needs. The demonstration demonstrated that the system can converse in multiple languages about complex subjects like Bengali poetry, Bollywood movies, and inventive masala dosa recipes.

According to Aggarwal, Indian language-specific models have outperformed open-source models of large languages. The time has come for us to fulfill Viksit Bharat's (advanced) dream and lead the AI vision for the next 25 years. Simulated intelligence will characterize the future worldview of economy and culture. One of the most significant technologies of the future will be artificial intelligence. Aggarwal stated, "We believe India must construct its own stack." The cultural and linguistic diversity of India is very different from that of the West, which has an impact on how AI is built. "We can reach 1.4 billion people with an India-first cost structure," he stated.

Their framework, he said, has been prepared on datasets that are exclusive and those got from accomplices.

Although Aggarwal did not disclose the pricing, he stated that it is a family of models with distinct price-performance balances. Speech models are also available alongside these language models. Thus, you can converse with the artificial intelligence models," he added.

In addition, Aggarwal has ambitious plans to construct their own AI chips within two years. We already have a sufficient amount of GPU resources. At last we will construct our own simulated intelligence supercomputer and our silicon. Additionally, we are developing our own data center technology; "To reduce costs, it is very important to make it more efficient," he stated.

Catch Daily Highlights In Your Email

* indicates required

Post Top Ad