“India should go for the full-stack approach in Artificial Intelligence,” Open AI’s CEO Sam Altman said in an interview with Hindustan Times on Wednesday, and reiterated this during a discussion with India’s IT minister Ashwini Vaishnaw later the same morning.
Altman’s comment in terms of what India should be doing is different from one he made in 2023 during a visit to the country. In response to a question, he had said at the time: “It is totally hopeless to compete with us on training foundation models,” and suggested that the energies and monies of the government and businesses could be better deployed elsewhere. “That was in a different context,” he said in the interview on Wednesday. “That was a time when frontier models were super-expensive to do.”
Things have changed since, he added, referring to the AI equivalent of Moore’s law – which says the processing power of a chip doubles every 18 months. “If you look at the cost curve of AI, we have been able to reduce the cost of a given level of intelligence about 10x every 12 months, which is unbelievably more powerful than Moore’s law.”
During the discussion, Vaishnaw, who said India’s driving philosophy with its AI program, as with all others, is to “democratise technology” said that the country is working on the “entire spectrum”: “Designing GPU chips; building the foundational models; and then building apps on top of that.” And who knows, he added, just like India completed its Moon mission at a fraction of the cost taken by other moon missions, maybe the country would succeed in creating “a (foundational or frontier) model at a fraction of the cost.”
{{/usCountry}}During the discussion, Vaishnaw, who said India’s driving philosophy with its AI program, as with all others, is to “democratise technology” said that the country is working on the “entire spectrum”: “Designing GPU chips; building the foundational models; and then building apps on top of that.” And who knows, he added, just like India completed its Moon mission at a fraction of the cost taken by other moon missions, maybe the country would succeed in creating “a (foundational or frontier) model at a fraction of the cost.”
{{/usCountry}}Foundational models, which form the basis of AI systems such as ChatGPT, are algorithms that train on vast amounts of data to develop broad capabilities in language understanding and generation. Vaishnaw said last week that six Indian entities had already shown promise and could release their foundation models in the next 10 months.
Altman told the audience at the discussion, which comprised developers, investors, startup founders and entrepreneurs (he also met separately later with a small group of them), that the “AI models on the near-term horizon are beyond the threshold”, and said they were ready for apps to be built on them. AI, he added, referencing Deep Research, the offering the company unveiled over the weekend, is “still in research assistant phase” and wasn’t “an innovator yet”. “I don’t think any of these models can go out and cure cancer,” he said, but they could help researchers do what they do better and “double the efficiency of every scientist on Earth.”
“India will build great models,” Altman said in the interview, adding that while the company did not have anything to announce about an India-specific investment, he hoped to have something “exciting to share soon”. India is Open AI’s second biggest market after the US, Altman said, without giving out specific numbers, “and users here have tripled in the last year”.
But while there is a “national game” to be played in AI, Open AI’s mission “which we take super seriously”, Altman said, “is for Artificial General Intelligence to benefit all of humanity”.
“I think this is one of these rare things that transcends national borders,” he added. “AI is like the wheel and the fire and the Industrial Revolution and the agricultural revolution. That is not a country thing. It belongs to everybody.”