Opening Remarks
Thank you. Firstly, thank you so much, Prithu. My name is Dr. Dibyadyuti Roy. I am an Associate Professor at the University of Leeds. I am deeply grateful to Prithu, the India Foundation for Arts, Dr. Bathwal, Professor Bibhuti, and many friends present here. A special acknowledgement to Rithvika from IFA.
Since this event launches Platform Bengali, I will briefly introduce myself in Bengali before returning to English. Given that IFA connects across multiple linguistic publics, I will continue in English for accessibility.
Framing the Talk
Today's talk is titled "Designs Taken for Wonders: Provocations on AI, Language, and Authority." My research examines cultural mediations of dominant and emergent technologies—their genealogies, narratives, and legacies. My academic trajectory began with nuclear cultures, moved into digital spaces, and later engaged with race, gender, and colonialism in digital environments. For the past six to seven years, my work has focused on social media, digitality, and algorithmic cultures.
Digital Humanities and Institutional Context
I have been fortunate to be a founding member of India's first digital humanities collective, Bharati (formerly DHAI). I was involved in organizing its first conference at IIM Indore in collaboration with IIT Indore. At the University of Leeds, I am part of the Digital Creativity and Cultures Hub, a multi-faculty centre where digital humanities is situated. I also run a series titled Deliberations in Digitality, which foregrounds voices from the majority world.
From Digital Humanities to Critical AI Studies
Over the past several years, my work has increasingly focused on what we call Critical AI Studies—particularly from a majority world perspective. The term "majority world," as many of you may know, was proposed by Shahidul Alam to challenge terms like "Global South" or "Third World," emphasizing instead where the majority of humanity resides.
Opening Provocation: A Question
On your screen are two events separated by approximately 250 years: On the left, the Mechanical Turk (18th century), presented as an automated chess-playing machine. On the right, Amazon's "Just Walk Out" stores (2023), marketed as frictionless AI retail. What is common between them?
What is common is that there is nothing artificial about artificial intelligence. The Mechanical Turk was later revealed to contain a hidden human operator. Similarly, Amazon's "Just Walk Out" system relied on human workers—reportedly in Gurgaon—monitoring CCTV feeds to enable the illusion of automation.
To paraphrase a well-known line: the greatest trick AI has ever pulled is convincing the world that it exists.
AI as Myth and Signifier
We live in a moment where everything is branded as AI—AI computers, AI classrooms, AI systems. This proliferation reflects not just technological development, but a discursive inflation. Drawing from linguistic theory, we might say that "AI" functions as a signifier whose signified has become unstable—overloaded with utopian and dystopian imaginaries.
Generative AI and Its Limits
What we currently call AI is largely generative AI, built on large language models. However, this is only a small subset within a larger hierarchy: Generative AI → Deep Learning → Machine Learning → Artificial Intelligence. At present, we have accessed only a very small fraction—perhaps 5%—of what might be considered human cognitive capacity, largely limited to object recognition and probabilistic inference. These systems do not "understand" in any meaningful sense. They operate through statistical pattern recognition and optimization.
The Political Economy of AI
Beyond utopian and dystopian narratives lies the material infrastructure of AI: data annotation, computational labour, global supply chains of digital work. Much of this labour is located in the majority world. The trajectory from sweatshops to call centres to data annotation reflects continuity in global labour hierarchies.
Histories of Artificial Intelligence
The term "artificial intelligence" was coined in 1956 by John McCarthy as part of a grant proposal—an "attractive term" designed to secure funding. AI has since undergone multiple hype cycles: the Cold War era (machine translation), the neural networks revival (1980s), and contemporary deep learning and generative AI. A key moment in recent AI history is 2012, marked by the success of deep learning models such as AlexNet, built on datasets like ImageNet developed by Fei-Fei Li.
Critical AI Studies
Critical AI Studies approaches AI as a socio-technical assemblage rather than a singular entity. It examines cultural meanings, ideological formations, material infrastructures, and power relations. AI can be understood as comprising three key components: training data, learning algorithms, and model applications. Each of these layers introduces potential for error, bias, and distortion.
Majority World Perspectives
Critical AI discourse differs across contexts. In the Global North: bias, fairness, accountability, transparency. In the Majority World: data colonialism, indigenous data sovereignty, postcolonial computing, feminist design practices. These approaches are complementary and must be brought into dialogue.
Conclusion
The future of AI lies not in abstraction, but in its contextualization within human worlds. Projects like Platform Bengali (BhashaScope) are crucial because they extend critical inquiry beyond academic spaces into community participation.
To paraphrase William Gibson: the future is already here—it is just not evenly distributed. Our task is to participate in redistributing that future.
