While there’s a primal, grassroots-level hunger around acquiring in data science knowledge in India, the country’s relative gap at the top of its talent stack is quite stark. India counts for just 386 of the roughly 22,000 Ph.D-educated AI-researchers worldwide, according to the Global AI Talent Report 2018 that crawled data from jobs site LinkedIn for its findings.

At tenth place in terms of countries with the most number of AI researchers, India trails the US (9,010), UK (1,861), Canada (1,154), France (797), Australia (657), Germany (626), Spain (606), Switzerland (441), and China (413).

The research, published by Jean-François Gagné, CEO of AI solutions provider Element AI earlier this month, notes that the representation may be skewed due to inherent biases in the dataset — for instance, LinkedIn’s relatively lower popularity in Asian markets. That shouldn’t apply in the case of India, though — Linkedin’s second-largest market by listings and the fastest growing. Judging by this data skew, which affects China’s standing in particular, it’s safe to say that India has the lowest PhD-level AI researchers per capita among major world economies. The silver lining? India is in second place among BRIC countries, ahead of Brazil (265) and Russia (88).

“We’ve typically seen that at the moment there’s around 7,000-10,000 data scientists in India, and we noticed that about 2% are PhDs,” says Rishabh Kaul, co-founder, Belong, a Bengaluru-based HR-tech startup. “There’s obviously a significant overlap between the way they have defined AI researchers and what we would consider as a data scientist.”

Others agree. “That seems like a very high number. I’m not sure how many of them would be doing quality work. I would say there are not more than 100, who will have impact, in terms of quality scientific publications or in industry,” says Vijay Gabale, co-founder and CTO at Bengaluru-based Infilect Technologies, the company behind Huew, an AI-based personal fashion assistant. He names professor Ravindran Balaraman from IIT-Madras and Soumen Chakrabarti from IIT-Bombay, as leading academics who have been doing good work, and publishing at top-tier conferences.

“Most of the quality AI research work comes out of eight to ten groups, who are active,” says Gabale. “So definitely in terms of comparison of activity that is happening in China, US, UK, or even France, we are not even in the race. I would say in terms of number of quality publications or tech innovations, we are not there yet.”

The really long haul

The talent report also looks at the number of conferences presenters in each country, as an additional data point, where India fares poorly. At thirteenth place globally, India ranks below the Netherlands and Italy on this metric.

These low numbers could undermine the country’s ambitions to “Make Artificial Intelligence in India and Make Artificial Intelligence work for India,” as expressed by Prime Minister Narendra Modi earlier this month at the inauguration of Wadhwani Institute of Artificial Intelligence in Mumbai.

One needs to sow the seeds decades before you’re ready to tell the world your ambitions, says Ashwini Asokan, co-founder of computer vision startup Mad Street Den with a presence in both the US and India. India’s tech industry and academia has not participated in or driven technology research and AI on scale, in any meaningful form, though she believes a handful of people in the private industry will forge ahead to build massive businesses that give Indians some hope.

“India risks becoming a spectator in the race to the future, we’re just starting to take note and have conversations about what it means to compete in a world that’s been thinking about this for decades,” says Asokan. “There will be implications for the lack of academic talent, in the short term. To make any meaningful progress, we first need to build schools of excellence, an approach and curriculum shaping towards a set of goals, globally qualified teaching talent and then build a culture of research and inquiry.”

India is placed nineteenth place based on the H Index (a metric that quantifies a country scientific productivity and scientific impact) on the Scopus data set, an abstract and citation database managed by information and analytics company Elsevier. The data, however, does indicate a jump in terms of numbers papers published — from 331 papers published in 2006 to 3,301 papers published in 2016.

Researchers like to flock together, so a company which has Ph.Ds with credentials is more likely to attract other similar high-calibre professionals, says Belong’s Kaul. Not having Ph.D-level researchers creates a bottleneck, as they work on breakthroughs and excite more people into academia. “Companies that deal with large amounts of data sets and serious computing power will also be able to attract good Ph.D folks,” says Kaul. The difference between academia and industry at least in AI and data science is reducing. Corporates have equally good, if not better tech infrastructure, he says.

Moonshots, not Ph.Ds

Then, there are those who don’t think much of the fewer Ph.Ds in the space. The number of AI researchers with Ph.Ds is not a very good measure of the country’s ability to adopt AI, insists V Kamakoti, a computer science and engineering professor at IIT-Madras, who recently chaired a recent government AI task force. “AI is people, process, and then technology. 70% of the work has to be done by subject-matter experts,” he says.

“Every problem doesn’t require a hardcore AI expert to guide it. Many times, certain templates will work. Whenever it doesn’t work, and you need a lot more insight into it… that’s when you need AI specialists,” he says. He named three of his colleagues — Professor Balaraman Ravindran, Mitesh Khapra, and Harish Guruprasad as people working on high-end AI problems from his campus.

“What cryptography means to cybersecurity, is what AI means to automation. Without cryptography, there’s no cybersecurity,” Kamakoti says. “But, nevertheless, if you take a cybersecurity process, cryptography will be a small aspect of it. It’s the same with AI.”

How about a moonshot contest like the DARPA Grand Challenge to galvanise Indians into an AI project of impact? The DARPA challenge, for example put the US at the forefront of the self-driving car industry, spurring research in this space in top US computer science universities by offering a prize of $1 million. Leading innovators in this space – such as Google X Lab founder Sebastian Thrun, and Otto co-founder Anthony Levandowski emerged out of this competition.

“During a discussion at the Rashtrapati Bhavan recently, the following question came up — what is the moonshot project of our generation today?” says Partha Pratim Talukdar, assistant professor at the Department of Computational and Data Sciences at Indian Institute of Science, Bangalore. “A problem that is sufficiently challenging that cannot be done over a weekend at a hackathon. A place where all the data science teaching could be utilised.”

Talukdar offers an answer — one that could address the significant literacy problem in the country and ensure that the less-literate don’t miss the benefits of the digital revolution. Speech interfaces could be one way to overcome this divide. But in most Indian languages, we don’t have good speech recognition technology, he says. “One potential moonshot could be that the government (and other interested parties) which have speech archives for different Indian languages,” says Talukdar. “What if you could make this available, to create, for example, an English-Assamese speech-to-speech translation service with 1% recognition word error rate in a year.” he says. “If you make problems like these, which have some utility, with a well-defined metric, a lot of people will get excited and contribute towards it.”

---