India plans 10,000-GPU autonomous AI supercomputer • The Register

WhatsApp Group Join Now
Telegram Group Join Now
Instagram Group Join Now

The Government of India has approved a ₹ 10,300 crore ($1.24 billion) funding package to bolster the country’s AI infrastructure.

The cornerstone of this effort is a planned supercomputer that will have at least 10,000 GPUs. The government has not released any other details of the machine – which will be part of the “India AI Compute Capability” – but has said it expects to require a public-private partnership to build the machine.

Another initiative will see the creation of a new academic institute: the “India AI Innovation Centre”, tasked with leading the development and deployment of core models. It is expected to place particular emphasis on local large multimodal models (LMMs) and domain-specific models. The center will focus on “leveraging edge and distributed computing” for maximum efficiency.

Funds will also go towards three other initiatives:

  • India AI Startup Financing Mechanism, which will streamline funding for both startups and industry-led AI projects to accelerate commercialization.
  • The India AI Datasets Platform, which will get more cash to improve public sector datasets so that local AI organizations and the government have the data they need to build appropriate AI apps.
  • the IndiaAI FutureSkills programme, which will improve access to graduate and post-graduate AI programmes, and establish data and AI labs that run basic AI courses in data and AI across India – especially outside major cities;

Two of the goals of the funding package are to “promote technical self-reliance” and “democratize the benefits of AI across all segments of society.”

It is unclear whether the planned supercomputer will accomplish these goals using domestic technology. While India has set itself the target of developing server-grade CPUs based on RISC-V architecture, Register So far there is no evidence of such devices being developed. And India is nowhere on GPUs.

There will be a push for local LLMs, though, because India has 22 scheduled languages ​​that need to be promoted by law. While some of these languages ​​- such as Bengali, Marathi and Telugu – have more than 80 million speakers, others are much less spoken. AI giants cannot prioritize LLM development for ~35 million Malayalam or Punjabi speakers.

India clearly intends to do such a thing itself.

Another omission in India’s announcement is the type of private partners to accelerate local AI development. India has a difficult relationship with Big Tech – praising its local investment while tightly regulating it, and unabashedly creating public digital goods aimed at making it harder for tech companies to build monopolies. . ®

WhatsApp Group Join Now
Telegram Group Join Now
Instagram Group Join Now

Leave a Comment