2 weeks Ago By Vallari Sanzgiri
Explainer Briefly Slides Jensen Huang, NVIDIA CEO, pushed for India to become the pioneer of manufacturing and export of Artificial Intelligence (AI) on October 24, 2024. Speaking at the NVIDIA AI Summit, Huang said that India should focus on producing digital intelligence while other countries still remain focused on manufacturing systems and chips. “No one manufactures intelligence right now.
There’s a factory running behind ChatGPT. It’s a new kind of factory..
. That’s a new manufacturing industry, don’t wait to jump into this one. Enter this one at ground zero,” said Huang, arguing that the export of AI would trump India’s export of labour.
NVIDIA announced partnership with Reliance Huang and Reliance Industries Chairman Mukesh Ambani announced a partnership between the two companies to build AI infrastructure in India. He said the ultimate goal with this partnership is to create that into a “fly-wheel” to create multiple AI models and export it. Largely the partnership will focus on three aspects, building infrastructure for computation, creation of an innovation centre in the country to provide NVIDIA technology to Reliance engineers and the creation of an AI application layer.
NVIDIA to work on a Hindi LLM Huang discussed plans to create a Hindi LLM for India in partnership with other Indian companies. This LLM will be open sourced and will be treated as an operating system for India, said Huang. However, it added that while NVIDIA will look into the building of the infrastructural technology and AI algorithm, the applications for the local market will be created by the companies.
“Hindi is the hardest language model region in the world. If anybody can do it, we can do it. Once India figures out a Hindi LLM, we can figure out LLM for everywhere [in the world],” said Huang.
Huang airs idea of Physical AI: Huang talked about the need for Physical AI for the next generation of AI technology. This is an AI that understands the physical world and functions effectively within the real world, be it an automated car, factory, etc. To achieve this, Huang called for an omniverse, a virtual world imitating the real world that trains and refines an AI model to function in the physical world.
AI won’t take your job, the one using it will When asked about AI’s limitations in carrying out human work, Huang said that AI cannot perform any human task at 100% and pushed for the technology to be viewed as an assistant. “It turns out that AI has no possibility of doing all that we [humans] do. Depending on the job, an AI can do 20% of the work a thousand times better.
However, in no situation can it do 100% of the job. So, AI should right now be used as an assistant. The person who uses AI to automate their 20% job will take your job,” said Huang.
AI to become cost-effective in coming years When asked how NVIDIA will make its products cost-effective and accessible to markets, Huang said that developing new AI tehcnologies will bring down costs by a hundred times every year. According to Huang, this is possible either by distillation techniques wherein a larger AI model will train a smaller AI model to carry out certain tasks in a cost-effective manner. While this smaller model will not work with the same agility as the larger model, it will address the issue of cost to some extent, said Huang.
Also Read:.
