No books, no human intervention, no training data and yet it learns, from first principles. Google-owned DeepMind showed last week that its artificial intelligence (AI) programme Alpha Go Zero was able to learn what knowledge was from itself. That’s spooky. Comparisons of AI to the invention of the steam engine or printing press for impact on humans now seem understated.

Even though a lot of AI expertise resides in the top 10-12 tech companies, governments are waking up to it as if it were an arms race. It may well be. In mid-October, the United Kingdom came out with its AI strategy. This was barely two months after China sprung its ambitious 15-year AI roadmap on the world. Last week, Taiwan said it’d spend $133 million on AI training. UAE just announced a minister for AI, whatever that may symbolise.

India’s ministry of electronics and IT (Meity) woke up to these “emerging technologies” this month. A few weeks earlier, journalists received emails that a “taskforce on the Artificial Intelligence for economic development, constituted by the Government of India, has invited public opinions”. It was from the ministry of commerce. It must be noted here that asking common people to comment without any large scale awareness programme or a broad stated vision amounts to a travesty. AI means different things to different people.

Three recent calls to action in big data and engineering have turned out to be disappointingly muted. The AI call to action mustn’t fall into that category. AI has existed for a long time but its applications such as speech/image recognition and search are impressively successful only now. Still, it can’t be borrowed and deployed so easily.

Startups are starting to do experimentation. So, if startups, math folks and business people get together, more can be done. Thus, awareness of opportunities and locating the right people with right skills, and making use of them is an opportunity

Sriram Rajamani, MD, MSR India

“AI is not like bringing bullet trains from Japan. You need algorithms which are trained on data. And data has context. You need to train algorithms on Indian data,” says Nisheeth Vishnoi, a computer scientist at Ècole Polytechnique Fédérale De Lausanne in Switzerland. If you have a self-driving car, it may be trained extensively on data elsewhere, but that’s a different context. “Developing AI is a completely different ball game as compared to developing space, power or train tech.”

As the two ministries define their AI turf, it’s vital to remember that hundreds of millions of people in India are exposed to algorithms, which perhaps already influence human behaviour.

AUTHOR

Seema Singh

Seema has over two decades of experience in journalism. Before starting The Ken, Seema wrote “Myth Breaker: Kiran Mazumdar-Shaw and the Story of Indian Biotech”, published by HarperCollins in May 2016. Prior to that, she was a senior editor and bureau chief for Bangalore with Forbes India, and before that she wrote for Mint. Seema has written for numerous international publications like IEEE-Spectrum, New Scientist, Cell and Newsweek. Seema is a Knight Science Journalism Fellow from the Massachusetts Institute of Technology and a MacArthur Foundation Research Grantee.

View Full Profile

Sign up to our India edition to read this story instantly

To sign up, you’ll create an account that will give you access to a new free story published once a week and archive of 214+ previously published free stories from our India edition. You’ll also receive one email every morning from us introducing the day’s story.

If you’ve already signed up, just enter your email below or login using Facebook or Google.