ChatGPT parent Open AI has pledged its support to the India AI Mission through participation in the app development initiative (ADI) under the programme, said a senior company executive on Wednesday.
“OpenAI is committed to supporting the India AI Mission’s application development initiative to ensure that Indian developers can build on our models and deliver social benefit at scale. We really look forward to continuing the conversation with the ministry and gauging where we might be able to add the most value,” said Srinivas Narayanan, vice-president, Open AI.
Identified as one of the seven pillars of India AI Mission, the ADI is aimed at promoting AI applications in critical sectors by addressing problems from central ministries, state departments, and other institutions.
Speaking at the Global India AI summit in Delhi, Narayanan said the firm was keeping India in mind while making important decisions in the large language models (LLM) space.
“Entrepreneurs understand market gaps. They are building innovative products and tools like ChatGPT are helping them accelerate this in completely new ways. We’re reducing the cost of artificial intelligence, enabling developers to write codes and are helping them create completely conversational and natural interfaces to computing,” he added.
Narayanan also said that cost and language were the two most important factors for the company in India.
“Couple of things specific to the India context — language is obviously on top of the mind for us and cost. These are the two pieces of feedback we’ve received from the community over the past year. GPT-4 addresses a lot of this feedback,” he added.
He said the company improved tokenisation, because of which it was able to use three times fewer tokens in the latest model, for processing Indian languages.
Tokenisation is the process of breaking down a piece of information like a sentence or a paragraph, into individual words or “tokens.”
These tokens are the basic building blocks of language, and tokenisation helps computers understand and process human language by splitting it into simpler units.
/cdn.vox-cdn.com/uploads/chorus_asset/file/25516879/108_JCA0100_comp_v004.1266.jpg)
/cdn.vox-cdn.com/uploads/chorus_asset/file/25509911/dodge.jpg)
