Building a world-class AI research lab in India

Building a world-class AI research lab in India

February 17, 2019 | Author: Varun Aggarwal, co-founder of Aspiring Minds

During the last 10 years I wore many different hats. After dropping out of a computer science PhD program at MIT, I started Aspiring Minds. Here, we use technology to measure the skills of two million jobseekers every year and match them with jobs at Fortune 500 companies. In a separate role, I also led an AI research lab that develops AI-based products and publishes in top AI conferences. The group is among the finest applied AI labs in India and runs the most popular machine learning outreach in the country (www.ml-india.org).

 

Where India Stands

A couple of years ago, I felt strongly about the lack of research in India. This is at a time when China was skyrocketing ahead of other countries (and is now ranked 2nd in the world, just after the U.S.) India ranks 12th[1]. In fact, when I conducted research for a book I wrote on the research ecosystem in India, I found our count of productive researchers is a mere 27,500, seven times lower than China and 17 times lower than the U.S. Lack of research means fewer globally competitive companies. Unfortunately, none of our companies made it to the MIT Technology Review’s Smartest Companies list (2012-2016). China had 14 and US had 105.

It is the same story for AI. China and the United States are vying for the top spot. Our high-quality AI research papers are less than that produced by a single Chinese university. AI is disrupting the world on a scale similar to the industrial revolution by mechanizing ‘human intelligence’. A whole lot of tasks and services that humans did will be done very efficiently by machines at little to no operational cost. Countries with the best science and AI systems will excel by not depending on the demographic dividend alone.

Unfortunately, India neither has a critical mass of AI researchers, nor applied AI personnel who can help our industry be competitive. Our companies face the grave risk of being edged out by newcomers. AI also provides a great opportunity to solve India’s social issues in health, education, sanitation, etc. We will miss the chance to leapfrog our social issues without developing our AI capability in a big way.[2]

 

Solutions

India needs multiple private initiatives in AI. The public institutions are impeded with structural issues and will take ages for reform. To up our game, we need to take cues from initiatives such as the Allen Institute for Artificial Intelligence, a research institute founded by late Microsoft co-founder Paul Allen; US-based nonprofit OpenAI, dedicated to building safe AI and having its benefits “as widely and evenly distributed as possible”; Vector Institute in Canada, dedicated to “drive excellence and leadership in Canada’s knowledge, creation, and use of artificial intelligence (AI) to foster economic growth and improve the lives of Canadians,” and burgeoning innovation in China.

I am building one such institution – the finest AI and neuroscience research destination in India and one of the best in world. The institution will have scale – geographic and temporal accumulation of talent. The vision is to take it to 80+ researchers and 200+ students in five years. Only with scale can we begin to make a real impact on Indian industry and the social ecosystem and spur a critical mass of globally competitive startups. For India’s size, we need big innovation.

The good thing about AI is that it has matured to make an immediate impact on businesses and society. At the same time, it still has big, unsolved long-term questions whose solutions will change our world. To stand the test of time, research institutions need to involve themselves in both aspects. Researchers at a lab can build, say, a dirt-cleaning robot. One can build a decent robot in two to three years that cleans a building (structured environment) by sophisticated applied AI research. On the other hand, to clean streets, roads and alleys, one needs to solve several hard problems in computer vision, robotics and so on – which will take years. The researchers at an AI institute can tackle both!

Third, an AI institution needs to have a globally connected people strategy. It needs distinguished and visiting Fellows from across the world to make a dynamic institution with ideas flowing in and out, almost daily. It also needs world-class permanent researchers sourced from top institutions not only in the U.S., but across the world. Great talent is out there – it just needs to be harnessed. Last, it needs a stream of young and dynamic kids, straight out of undergraduate universities, who will not only be the workhorses, but bring in fresh ideas. They will impact the industry and society at large.

All of this has to go hand-in-hand with vigorous interaction with the ecosystem. Data is ‘air’ for AI – the institute has to forge data partnerships with the industry and social organizations to also source questions and deploy technology. The institute will have a dedicated Data Team to enable this, and also have a software development team for prototyping solutions, releasing APIs, and developing experimental benches. An entrepreneur in residence is a no-brainer!

All of this can happen and is happening! It can propel India’s academia and industry to an entirely new dimension – and as a result, our society. But it needs the support of the Indian diaspora in a big way – in terms of partnerships, advice, and funding.

This is an important mission for India.

 

Footnotes

[1] This is based on total citations of research papers produced by  the countries in the years 2012-2016. Please refer to Chapter 3 of the book, Leading Science and Technology: India Next? for more details.

[2] India’s mess of complexity is just what AI needs.

 

Varun Aggarwal is the co-founder of Aspiring Minds, one of the world’s largest skill assessment companies. He is also the author of the book Leading Science and Technology: India Next? (March 2018). He writes poems and stories for leisure.