This report is part of "A Blueprint for the Future of AI," a series from the Brookings Institution that analyzes the new challenges and potential policy solutions introduced by artificial intelligence and other emerging technologies.
Size of the AI sector in India
The size of the AI sector in India is difficult to determine, given that a lot of AI applications are in intermediary phases of production. Globally, one popular means of measuring the size of AI sectors is by adding up private sector investment in AI start-ups. According to one estimate, total AI funding worldwide has increased from $862 million in 2012 to $6.4 billion in 2017.1 The Indian AI sector, too, has seen growth in this period, with a total of $150 million invested in more than 400 companies over the past five years.2 Most of these investments have come in the last two years, when investment nearly doubled from $44 million in 2016 to $77 million in 2017.3
In India, too, the government is spearheading investments in AI and other emerging technologies. In the latest budget, the government set aside $480 million for investment into emerging technologies including AI. This commitment could help put India on the map, as this outlay compares favorably to those of Australia, Canada, and the European Union.5 However, it pales in comparison to the $7 billion minimum that China has committed to invest in AI through 2030.6
National AI strategy
In June 2018, the National Institution for Transforming India (NITI Aayog) published a discussion paper that outlines India’s national AI strategy. India is one of more than 20 countries to have released their national strategies in the last two years. Countries have chosen different facets of the AI sector, from research and development to national security, as national priorities.
India’s national strategy stands out for its focus on the social sector.7 The paper is an ambitious roadmap for the AI sector in India and identifies three ways in which AI can serve the country’s needs. First, AI will contribute to India’s economic growth by overcoming the physical limitations of capital and labor, to open new sources of value across the country. Second, AI applications in the social sector will serve to improve the quality of life of ordinary citizens. Third, India can serve as a garage or tinkering lab to develop new AI technologies for other developing countries.8
To this end, the NITI Aayog has identified five priority sectors where AI investments should be focused: health, education, agriculture, smart cities, and smart mobility. These sectors were identified based on successful interventions in other countries and their potential for high social impact within India. Agriculture is the exception in the international experience category, as there has been comparatively less focus on this sector globally. But given the importance of agriculture to India, the NITI Aayog has identified that sector as an area where India can play a lead role in implementing AI solutions.9
Given the importance of agriculture to India, the NITI Aayog has identified that sector as an area where India can play a lead role in implementing AI solutions.
The AI discussion paper also highlights the need for greater investment in AI education and research across the country. To this end, the NITI Aayog recommends establishing centers of excellence for AI research. An interesting facet of this strategy is the pursuit of ‘moonshot projects,’ or ambitious, exploratory, and ground-breaking projects that can help push the frontier of knowledge in AI.10 Moonshot projects have helped companies like Google and Amazon occupy a leadership role in AI over the last few years. There are fundamental constraints, however, that pose a challenge to this strategy in India. We point out a few of these challenges below, including the country’s quality of connectivity and education, which will determine the future successes of India’s AI ecosystem.
India is uniquely placed to take advantage of developments in the AI space. The NITI Aayog report goes into some detail in identifying these incumbent advantages. There are two facets of the Indian economy in particular that can help spur the growth of AI.
Prominent conglomerates looking to streamline operations
Despite the increase in funding for AI start-ups around the world, the number of companies receiving funding has fallen in 2017.11 The AI sector is known to favor more established players who have access to large data sets and capital.12 Since developing workable AI platforms is resource- and capital-intensive, there have been few commercial applications in the consumer space. So far, most commercial applications have focused on improving the efficiency of businesses.
This is also the case in India. Chatbots, which have been employed by many tech platforms as a replacement for their customer service operations, are the most popular use of AI that consumers experience. However, the most successful applications of AI have been on the business side. Mad Street Den is a good example of this trend. Their AI platform helps e-commerce platforms catalogue their products based on user behavior.13 This allows companies to target customers more effectively and to reduce costs. Insights from user behavior are also being used to help design better products.
A report by McKinsey & Company notes that profitability in the consumer space will likely be limited because enhanced user experience does not necessarily lead to increased revenues. With new technologies, consumers also tend to gravitate toward free services. For this reason, McKinsey notes that governments and businesses are most likely to purchase AI technologies to improve efficiency. Banks, for instance, can use AI to detect suspicious lending activity and money laundering.14 Given the grave concern of NPAs (non-performing assets) in the Indian banking sector, AI solutions could be adopted as “early warning systems” to raise red flags on poor repayment behavior, etc.
Enterprise solutions are also suitable to the Indian market given that the 30 largest conglomerates account for 40 percent of all sales in the country.15 Conglomerates also account for 56 percent of all assets in the country.16 At a time when conglomerates are looking to slim down,17 AI applications can help streamline their businesses. The challenge facing these companies, according to McKinsey, is to identify ‘micro-verticals’ where AI applications can help improve efficiency.
Large number of STEM graduates
India has a large number of graduates from technical programs. In 2017-18, close to two million students graduated with STEM (science, technology, engineering, and math) degrees in India.18 This means there is a large base of individuals who have the aptitude to work in the AI space. The NITI Aayog recommends that these students can be trained to work in new jobs created by AI, and to contribute to the development of AI research in India.
However, this ignores the twin problems of low-quality of higher education in India and the significant brain drain from the best institutions over the last several decades. The 2018 India Skills Report found that only 45 percent of graduates are employable. For engineering graduates, employability varies between 10 to 40 percent, depending on the role.19 To address the skills gap in fresh hires, many companies invest in lengthy training programs, which often retrace basic concepts that should have been taught in college.
An obvious solution to this problem would be the creation of training institutes to impart skills for an AI market. However, India also has a shortage of qualified faculty to teach AI courses. The government has launched many programs to improve the quality of teaching and to promote knowledge exchanges with universities abroad. These programs can be tweaked to focus on AI training.
The second problem is the ‘brain drain’ that happens from the best technical institutions across India to developed markets globally. Reversing this trend in order to retain talent in India will require a concerted strategy. India has much to learn from China in this regard.
Almost 80 percent of Chinese students who study abroad return to China after graduation. This trend is driven by government policies encouraging graduates to return and a domestic job market that is stronger than most developed countries. In the specific case of AI and other highly specialized fields, the government offers lucrative research grants to encourage experts to return to China. For instance, since 2008, more than 7,000 Chinese scientists have been awarded research grants of 2 million Yuan20 (approximately $300,000) along with housing benefits and cash rewards. In the AI sector, experts can expect much higher salaries, ranging from $300,000 to $400,000.21 Unlike in the past, China-based companies are able to compete with U.S. companies in paying top dollar for these experts. It was recently announced that China is investing $2 billion to set up an AI park in Beijing.22 There are also instances of private players setting up research centers, sometimes with government help. For example, a new AI school has been set up by a Google alumnus with many U.S. experts among its faculty.
Brain drain occurs from the best technical institutions across India to developed markets globally. Reversing this trend in order to retain talent in India will require a concerted strategy.
A fundamental challenge with regards to the development of AI in India is the quality of connectivity in the country. India has taken great strides in improving access to the internet over the last decade; the internet penetration rate reached 30 percent in 2016—a huge jump from less than 4 percent in 2006. However, there has not been a corresponding increase in the quality of internet access. The development of future AI technologies is closely linked to 5G networks.23 India is not expected to begin rolling out 5G networks until 2020, and it may take up to five years to fully deploy.24
Of the more than 300 million internet users in India, 77 percent of urban users and 92 percent of rural users access the internet through mobile phones as their primary device.25 This is not unique to India, as more than 50 percent of world users connect to the internet using mobile devices. However, internet speeds on mobile devices in India are much slower than global standards. For instance, a recent report by Open Signal showed that average 4G speeds for most operators in India (between 2-6 Mbps)26 are much lower than the global average (17.4 Mbps)27.
With limited access to high quality internet, most users are not able to utilize the full range of services available on the internet. A 2015 GSMA study showed that mobile devices in the Asia-Pacific region are usually used for entertainment (e.g., watching videos, streaming music).28 Due to limited bandwidth, there are fewer services that can improve productivity on mobile platforms. This has a direct impact on the development of AI in India.
Some connectivity issues can easily be overcome. For example, when deploying AI in support of precision agriculture, advisories can be sent to farmers via text message. However, for more complicated uses of AI, such as self-driving cars, decision-making must happen at near zero latency.29 Therefore, if India is to become a hub for the development of new AI technologies, it must invest to enhance the quality of its connectivity infrastructure.
AI ‘black box’
India’s AI strategy stands out for its overt focus on the social sector—but the implications of such use in the social sector are not fully clear. This is because of the AI ‘black box.’30 AI is different from other existing technologies in that it functions on training, rather than instructions. Given a data set, AI is trained to perform a particular task. How the AI performs the task is not always under the control of its users. This can lead to many negative real-world implications, as AIs can be biased by the data set they are exposed to.
One of the most famous examples of the AI black box malfunctioning is the Microsoft Chatbot. In a short amount of time, the bot was trained by third parties to tweet out racial slurs, and was quickly shut down.31 The larger concern with AI is that the programmer and user have very little control once the technology is deployed. Concern will only be amplified when AI is used to implement large-scale social schemes. China is already using AI to help administer its program of scoring citizens based on their behavior.32 It has been reported that AI has been used for surveillance of social media accounts, credit scores, and more as part of that program.
One response to the black box problem has been to demand greater transparency in algorithms. A movement toward fair algorithms also seeks to embed ethics into AI programming.33 Since the amount of human control in AI is limited, such a framework of ethics or policies in India should be established before the technology is used in the social sector. There is a need for more research on the implications of using AI for social programs.
Improve connectivity through BharatNet
The IT revolution of the 1990s and 2000s in India was possible due to government investment in electronics and computing in the preceding decades.34 Similarly, the development of AI needs to be supported by sufficient investment in rolling out high-speed internet across the country. In 2012, the Indian government launched the ambitious National Optical Fibre Network project (since renamed to BharatNet) to connect all 250,000 gram panchayats in India with high-speed internet. Originally slated to be completed by 2015, the project only recently completed its first phase of connecting 100,000 gram panchayats last year.35 As of August 2018, 114,348 gram panchayats have been connected.36
In its second phase, BharatNet has invited private sector participation in eight states to complete the project.37 However, even in many areas where the cable has been laid, connectivity is yet to become a reality. Among other reasons, this is because of a lack of clarity on the division of funding responsibilities between the center and state governments.38 Because it would be difficult for the government to fund free internet to connected villages on its own, it might be wiser to incentivise or partially subsidise private players to provide internet connectivity at affordable rates.
Create incentives to reverse the ‘brain drain’
As previously mentioned, brain drain of top-talent from India presents a significant problem; in 2016, 278,383 Indian students were pursuing tertiary education in other countries—almost double the number from a decade earlier.39 Indian students studying abroad accounted for 1 percent of India’s total enrollment.40 A clear majority of Indian students abroad are studying at the postgraduate level and in STEM disciplines. Overseas Indian students account for almost 7 percent of postgraduate enrollment in India. This is despite the fact that higher education in many of the top destination countries is far more expensive than in India.
In the context of our AI discussion, this means that it is very difficult for India to retain its best talent. India should look to replicate the Chinese model in bringing back expertise in strategically important fields like AI. The government may not be able to fund such incentive programs on its own. But as we discussed, large conglomerates have a lot to gain by implementing AI solutions. The government is already funding infrastructure investments in higher education through CSR funds.41 Extending this model to attract the best AI talent to the country would be a step in the right direction.
Training programs for AI
Even with private sector support, it would be difficult for India to match China in terms of investment in AI. For future growth in the sector however, India has a large base of students enrolled in STEM programs who can be trained to be AI professionals. So far, professional training programs for STEM graduates has been limited to those offered by technology companies who invest in training their campus recruits. However, this training is limited to the kind of projects that the respective companies work on. AI training should focus more broadly on equipping professionals with the necessary skills to work in a new AI economy.
The NITI Aayong has taken a longer-term view and recommended setting up ‘tinkering labs’ starting at the school level. The government is in the process of establishing more than 2,000 tinkering labs to teach students to engage with technology from a young age.42 This is a welcome move, but it does not address the relatively shorter-term need of a skilled workforce to work with AI and other emerging technologies.
The government is in the process of establishing more than 2,000 tinkering labs to teach students to engage with technology from a young age.
In this respect, we make two suggestions. First, colleges in India should use online courses to help bridge the gaps in expertise in AI. The few AI experts in India are limited to elite institutions like the Indian Institutes of Technology (IITs). More importantly, most colleges in India do not have postgraduate programs where AI expertise can be cultivated. However, many reputable online AI courses already exist, which can be included in the course curriculum.43
Second, colleges should look to foster closer ties with industry. Though the AI industry is small, it can benefit from connections with educational institutions and vice-versa. Research shows that the IT revolution was possible in the 1990s because there was an increase in the number of computer and electronics engineering colleges in hubs like Bangalore.44 This helped establish IT clusters. Similarly, AI can benefit from upgrading existing clusters or establishing new clusters around growing AI businesses. The technical education regulator, AICTE, recently mandated internships for all engineering students.45 Such programs can easily be extended to the emerging AI sector.
The issues outlined in this paper represent areas where investment will help create an ecosystem for AI in India. Just as investment in computing technologies in the 1970s and in technology clusters in the 80s and 90s spurred the IT revolution over the last two decades, implementing the above recommendations will help kick start the AI sector in India.
- Tracxn, “Artificial Intelligence Sector Report 2018”, February 2018, available at <https://tracxn.com/reports/fMfLAJmqwTD4z61mrkRwbUsWTnIETKriWGCELb>.
- NASSCOM, “Artificial Intelligence Primer”, at p. 16.
- Supra, Tracxn.
- CB Insights, “Top AI Trend to Watch in 2018”, available at < https://www.cbinsights.com/reports/CB-Insights_State-of-Artificial-Intelligence-2018.pdf>.
- Tim Dutton, “An Overview of National AI Strategies”, June 29th 2018, available at <https://medium.com/politics-ai/an-overview-of-national-ai-strategies-2a70ec6edfd>.
- Dave Gershgorn, “AI is the new space race. Here’s what the biggest countries are doing”, Quartz May 2nd 2018, available at <https://qz.com/1264673/ai-is-the-new-space-race-heres-what-the-biggest-countries-are-doing/>.
- Supra, Dutton.
- NITI Aayog, “National Strategy for Artificial Intelligence Discussion Paper”, June 2018, at pp. 18-19.
- Id, at p. 56.
- Supra, NASSCOM.
- Sriram Sharma, “Here is Why India is Likely to Lose the AI Race”, Factor Daily August 18th 2018, available at <https://factordaily.com/artificial-intelligence-india/>.
- See <https://www.moneycontrol.com/news/business/high-fashion-goes-tech-savvy-as-ai-takes-the-ramp-in-india-2807961.html>.
- Gaurav Batra, “Artificial Intelligence: The Time to Act is Now”, McKinsey, January 2018, available at <https://www.mckinsey.com/industries/advanced-electronics/our-insights/artificial-intelligence-the-time-to-act-is-now>.
- Aditya Mohan Jadhav and V Nagi Reddy, “Indian Business Groups and their Dominance in the Indian Economy”, 29 (2) EPW, July 2017.
- See <business-standard.com/article/companies/the-end-of-conglomerates-117031700943_1.html>.
- See <https://blogs.timesofindia.indiatimes.com/toi-edit-page/the-conglomerate-dilemma-indias-conglomerates-need-to-slim-down-to-reasonable-sizes-and-focus-on-a-few-core-businesses/>.
- All India Survey of Higher Education 2017-18.
- See for instance the Aspiring Minds Employability Report, available at <https://www.aspiringminds.com/sites/default/files/National%20Employability%20Report%20-%20Engineers%20Annual%20Report%202016.pdf>.
- See <https://www.scmp.com/tech/leaders-founders/article/2140624/home-now-seen-land-opportunity-chinese-tech-graduates-us>.
- CBInsights report, “Top AI Trends to Watch in 2018.”
- MIT Technology Review: <https://www.technologyreview.com/the-download/>.
- Tam Harbert, “How 5G will make AI-Powered Devices Smarter”, Intel, available at <https://iq.intel.com/how-5g-will-make-ai-powered-devices-smarter/>.
- See https://economictimes.indiatimes.com/industry/telecom/telecom-policy/5g-panel-suggests-opening-of-new-spectrum-bands/articleshow/65515228.cms.
- IAMAI, “Internet in India- 2016”, available at <http://bestmediainfo.com/wp-content/uploads/2017/03/Internet-in-India-2016.pdf>, at p. 6.
- Open Signal, “State of Mobile Networks: India” April 2018, available at <https://opensignal.com/reports/2018/04/india/state-of-the-mobile-network>.
- Akamai, “State of the Internet Connectivity Q1 2017 Report”, available at <https://www.akamai.com/fr/fr/multimedia/documents/state-of-the-internet/q1-2017-state-of-the-internet-connectivity-report.pdf>, at p. 28.
- GSMA, “Mobile Internet usage challenges in Asia - awareness, literacy and local content",July 2015, available at <https://www.gsma.com/mobilefordevelopment/wp-content/uploads/2015/07/150709-asia-local-content-final.pdf>, at p, 11.
- Latency is the amount of time taken for a packet of data to travel from one point to another.
- See for a discussion, https://www.technologyreview.com/s/604087/the-dark-secret-at-the-heart-of-ai/.
- See https://www.theguardian.com/technology/2016/mar/24/tay-microsofts-ai-chatbot-gets-a-crash-course-in-racism-from-twitter.
- Jack Karsten and Darrell West, “China’s Social Credit System Spreads to more Daily Transactions, Brookings June 18th, 2018, available at <”https://www.brookings.edu/blog/techtank/2018/06/18/chinas-social-credit-system-spreads-to-more-daily-transactions/>.
- In particular, there has been focus on setting standards for government use of algorithms, see https://www.nesta.org.uk/blog/10-principles-for-public-sector-use-of-algorithmic-decision-making/.
- For a discussion on Government policies leading up to the IT revolution, see V Rajaramana, “History of Computing in India (1955-2010)”, available at <http://www.cbi.umn.edu/hostedpublications/pdf/Rajaraman_HistComputingIndia.pdf>, pp. 18-39.
- See <https://economictimes.indiatimes.com/industry/telecom/telecom-policy/bharatnet-has-a-new-target-connect-every-village-home/articleshow/64212095.cms>
- See <http://bbnl.nic.in/index1.aspx?lsid=570&lev=2&lid=467&langid=1>
- See <https://www.dailypioneer.com/state-editions/dehradun/cgarh-has-own-model-for-bharatnet-implementation.html>
- See for a discussion <https://www.thehindubusinessline.com/specials/india-file/the-broadband-tangle/article22859050.ece>
- UNESCO UIS database.
- Also known as the Outbound Mobility Ratio.
- The Higher Education Funding Agency is a newly created body that raises private capital from CSR funds to fund invest in higher education infrastructure at elite institutions. See <http://pib.nic.in/newsite/PrintRelease.aspx?relid=173962>.
- See <http://aim.gov.in/atal-tinkering-labs.php>.
- See for instance, Andrew Ng’s 5 course AI introduction programme <https://www.coursera.org/specializations/deep-learning>.
- See for a discussion, MH Bala Subrahmanya, “How did Bangalore Emerge as a Global Hub of Tech Start-ups in India? Entrepreneurial Ecosystem —Evolution, Structure and Role”, Journal of Developmental Entrepreneurship Vol. 22, No. 1 (2017), available at <https://www.worldscientific.com/doi/pdfplus/10.1142/S1084946717500066>.
- The AICTE is also working with online platforms to make it easier for students to get internships. See <https://www.aicte-india.org/downloads/MOU%20Internshala.pdf>.