'Govts should access the best tech, not reinvent the wheel'

Countries aspiring to build large language models from scratch may risk falling behind: Nadella

- Leslie D'Monte

In May, Sam Altman, co-founder and chief executive officer of Microsoft-backed OpenAI who was fired on Friday, had stirred a hornet’s nest, commenting on India’s attempt to develop a generative artificial intelligence (AI) tool like ChatGPT, terming it “hopeless", while responding to a question by Rajan Anandan, the managing director of Peak XV Partners, and a former head of Google India. Altman later clarified that his response was “…really taken out of context! (and that the)...question was about competing with us with $10 million (as opposed to $100 million), which I do think is not going to work..."

That said, India does not have a semiconductor fab to build chips required to power a large language model (LLM) or LLM-powered chatbot. LLMs typically require below 10 nanometre (nm) graphic processing units (GPUs), while India plans to build a 40 nm semiconductor fab, implying that it will have to buy graphics processing units (GPUs) and use cloud services to bear the costs of building an LLM.