AI Narratives in India: Workshop Summary and Commentary
On March 11-12 2021, hundreds of people tuned in to the virtual Global AI Narratives: India workshop, a collaboration between researchers at the Leverhulme Centre for the Future of Intelligence, University of Cambridge and the School of Arts and Sciences, Ahmedabad University.
The two-day workshop covered an array of topics ranging from the representation of AI in Indian science fiction (panel 1) and the diverse linguistic literary traditions in India (panel 4), to the history and philosophy of AI in India (panel 2) and the controversial implementation of the Aadhaar biometric identification program (panel 3). In this blog we discuss highlights of each panel discussion. Full video recordings of the event can be viewed here.
AI in Indian science fiction and imagination - Panels 1 and 4
The opening panel aimed to explore the representation of intelligent machines in Indian science fiction in presentations by Samit Basu (author of The Gameworld Trilogy and Chosen Spirits), Divya B. (author of Machinehood) and Tanuj Solanki (author of The Machine is Learning). Science fiction is an important topic of conversation because it can be a key driver of public imagination, expectations, hopes, and fears regarding speculative futures and emerging technologies. The genre provides a platform for preemptively engaging with difficult ethical questions and exploring anxieties. It is also noteworthy that the influence between technological advances and science fiction is not unidirectional. Popular science fiction can also influence the direction of future technological development.
Samit Basu noted examples of intelligent machines that can be found in Indian literature. For instance, Basu recalled a story about the Buddha's tomb being defended by an army of robot warriors with which various heros fought in an attempt to rescue the body. However, while such stories exist, Basu explained that only readers with a particular interest take the trouble to seek them out. It would be inaccurate to say that there exists a tradition of imagining intelligent machines in Indian literatures.
Perhaps the most interesting takeaway from the science fiction panel was that it is difficult to discuss representations of AI in Indian science fiction precisely because it is difficult to characterise Indian science fiction. Basu noted that while the science fiction genre certainly exists in India, it is represented in a plethora of different languages. Most people in India are polyglots, but the country’s linguistic diversity poses a hurdle to establishing a central literary core.
Furthermore, there is not much storytelling about AI in Indian languages because the central core of storytelling in India is conformist and centralised. This point was reiterated by all three panelists. As Divya B. explained, readers in India appreciate modern science fiction, but they do not look for it in the Indian marketplace. India has a long history of literature in mythology and fantasy, and these genres dominate over science fiction.
Basu also pointed to a knock-on effect resulting in the absence of science fiction in Indian cinema. Science fiction films are expensive to produce, Basu explained, and so film producers opt for a risk averse strategy of retelling classic stories that are proven to sell to wide audiences. Consequently, most science fiction cinema in India is imported from the West.
The final panel of GAIN: India, ‘Responses to AI and Science Fact-Fiction in Hindi, Bengali, and Malayalam Literature’, further explored India’s diverse literary traditions and their effect on modern science fact-fiction in much more depth. For those interested, we have composed a second blog post dedicated to the topic.
Despite the shortage of science fiction in Indian literatures, the panellists still spoke to what a distinctly Indian interpretation of AI and intelligent machines would look like. Basu argued that the representation of “AI in science fiction boils down to questions of power and inequality… Who gets to use it, who gets to design it, who gets to have how much control over where it is used?” Tanuj Solanki expanded on Basu’s point by speaking about fear of job loss and social class divisions and inequality. Solanki described AI-enabled technologies as potential tools for reinforcing the social divisions that already exist in South Asia and explained how the theme of tech subordinating social divisions creates the kinds of dystopian scenarios on which science fiction writers thrive.
History and philosophy of AI in India - Panel 2
The second panel took a step back to look more broadly at the history and philosophy of AI in the subcontinent. Each speaker took a very different approach to addressing the topic. The panel demonstrated the many ways the history of AI in India can be studied and used to better understand India’s role in and experience with AI development.
First, Professor Daud Ali (South Asian Studies Department, University of Pennsylvania) presented on the history of automata in medieval India (10th-13th century). Ali described how animated machines in the forms of humans, animals and plants were installed in menagerie gardens as objects of beauty and wonderment. The purpose of these machines was not to aid in the mastery of nature or to deceive onlookers into believing they were real. Rather, they were meant to bring pleasure derived from the observer’s awe and fascination with how the artisan had been able to masterfully conceal the automata’s mechanical nature. The third panel speaker, Dr. Noopur Raval (AI Now Institute, NYU), noted that Ali’s historical account of automata shows there is room for dialogue about intelligent machines couched in terms of wonderment and beauty as opposed to moral crisis, as is most common in the West.
Moving forward several centuries, Arun Sukumar (author of Midnight’s Machines: A Political History of Technology in India) discussed India’s current role in influencing the direction of AI development. He explained why India has little influence on AI inventiveness and what would have to change for this to be different.
First, Sukumar explained that AI research and development has flourished in Silicon Valley because of its fast-moving, anti-establishment culture represented by its most striking mantra, “fail fast, fail often”. In contrast, India is a more risk-averse society with a strong social stigma against failure. Such conditions are not conducive to entrepreneurial explorations into a cutting-edge technology.
Secondly, Sukumar noted that India does play a key role in AI development by providing essential data labeling services. Pictures, videos, sound clips, etc. have to be labeled by humans before they can be used to train machine learning models. However, decisions about how the data is labeled are made by the parent companies which are primarily located in the West. Consequently it is unlikely that uniquely Indian perspectives would influence AI development at this stage. As panelist Noopur Raval commented, Indian data labelers are expected to engage in the practice of erasing their own cultural etchings.
Overall, Sukumar argued that so long as India contributes to AI development on the margins, the country will not have much effect on the technology’s development trajectory. Indian cultural values will only come in to play if adopted by individuals higher up the chain.
Noopur Raval closed the session by bridging the past and the present in her discussion of how India's colonial past must be taken into account in order to develop fair and beneficial AI. Raval began by pointing to recent discussions in the United States about how AI systems trained on historical data expound on racial biases that stem from the country’s not-too-distant history of slavery and racial segregation. Raval argued that any demographic data on which AI systems are trained in India will similarly bear the marks of the country’s colonial, classist past if effort is not made to counter the effect.
Perspectives on UIDAI and Aadhaar - Panel 3
The third workshop panel addressed India’s recently rolled out biometric identification program, Aadhaar. An Aadhaar number is a unique 12-digit number issued by the Unique Identification Authority of India (UIDAI) that links to a person’s biometric identity (for example using iris scans or fingerprints). Proof of identity is necessary to access a variety of public resources and services, since many Indian residents, especially in more rural areas, lack proper documentation. The aim of the Aadhaar program was to provide every Indian resident with a unique and secure proof of identity.
The first two panel speakers, Srikanth Nadhamuni and Ram Sewak Sharma, along with Nandan Nilekani (not present) led the team that developed the technology and architecture that undergirds Aadhaar and oversaw the enrollment of India’s 1.3 billion residents.
Together, Nadhamuni and Sharma shared the journey of how Aadhaar began as a comparatively simple project of embedding unique identification numbers into birth certificates that evolved into what the UIDAI is in its current form. Sharma discussed the unique design challenges to building a universal identification system in India from the ground up: how do you create identities in places where people have no documents? This challenge along with requirements for guaranteed uniqueness of issued identities and mass scalability (India has the world’s second largest population) incited the switch to a biometric approach.
Nadhamuni spoke further about Aadhaar’s design and development. Aadhaar started as a Silicon Valley-style “garage startup”. The Aadhaar project was spearheaded by a small team out of an apartment in Bangalore with the project leads doing much of the design and development work themselves.
Nadhamuni noted that most governments would opt to outsource such a project to large multinational corporations; however, Aadhaar’s small group had the advantage of designing the system from scratch to accommodate India’s unique and unusual requirements.
Nadhamuni concluded his presentation by commenting that until now, India’s tech scene has largely been viewed as developing software solutions for export to the rest of the world. However, the success of the Aadhaar project has given India a new confidence that it can embark on its own technologically driven digital transformations. Building on this reinforced confidence, Nadhamuni expressed his optimism about India’s ability to adopt and develop AI-enabled solutions as well.
For more information on the design and development of Aadhaar, see panelist Ram Sewak Sharma’s recent book, The Making of Aadhaar (2020).
The Aadhaar project is a feat of technological development and scalability, but the final panel speaker, Dr. Chinmayi Arun (Fellow at the Information Society Project, Yale Law School) embarked on a more cautionary discussion of how techno-political systems like UIDAI interact with society. Arun noted that while the small Aadhaar team facilitated the technology’s rapid development and deployment, the project moved too quickly to ensure legal and social legitimacy. In scientific institutions, structures are built in to ensure proper epistemic and ethical standards are upheld. If researchers wish to embark on higher-risk projects with the potential for harm to individuals or society, they have to jump through significant regulatory hoops to gain approval. Aadhaar, on the other hand, had little or no bureaucratic oversight of this sort. Accordingly, even though Aadhaar has the government’s approval, many people are unconvinced about the program’s benefits and about the government’s foresight in predicting and taking steps to prevent potential harms.
Chinmayi Arun also commented that while technological and bureaucratic experts engaged in significant reflections and discussions about the potential impacts and implementation challenges of Aadhaar, given its potentially wide-reaching social impacts, the seeming absence of human rights and social impact experts was disappointing. Arun emphasised the importance of engaging more with professionals like anthropologists and political scientists to find humanity in such projects.
Conclusion
The AI Narratives in India workshop covered a wide range of topics, from historical perspectives on autonomous machines, to AI representations (or lack thereof) in contemporary Indian science fiction, to the modern-day environment of tech development and adoption in India. It is our sincere hope that these conversations will continue as we strive to nurture a global intercultural dialogue on the hopes, fears, and perceptions of emerging technologies.