In the digital age, encyclopaedic knowledge has become both a battleground and a commodity. It is the story of a basic public utility, our shared memory, slipping through our fingers.
Mumbai: Last month, as a nation debated whether its new Parliament building should be photographed with or without the national flag, two teenagers in different corners of India went hunting for a quick answer. One opened Wikipedia on a cracked phone screen; the other asked Grok, the chatbot inside Elon Musk’s social feed. Both wanted the same thing: a short, reliable note on the building’s architectural symbolism. Within minutes, each had a paragraph. Neither realised that the paragraphs quietly disagreed on who first proposed the lotus motif. Wikipedia credited a French-trained Bengali engineer; Grok named a British planner from 1911. Both sounded equally confident. Both supplied links—Wikipedia to a 1952 monograph, Grok to a 2023 blog post. Somewhere between the two, the plain idea of “what happened” dissolved into competing fumes.
In the digital age, encyclopaedic knowledge has become both a battleground and a commodity. This is no longer a story about two websites. It is the story of a basic public utility—our shared memory—slipping through our fingers. For twenty years, we could blame Wikipedia’s skewed lens: too many male Western editors, too few Tamil citations, too much weight given to headlines from Delhi and London. Today, we are promised a fix. Grok’s creators say their AI will scan everything and deliver the neutral residue. Yet, early evidence shows the fix is simply a new filter, tinted a different shade. If we keep bouncing between these two flawed atlases, we will keep getting lost. What we need is a third reference, built from scratch: transparent, traceable, and so boringly even-handed that no single culture can claim proprietorship over it.
Wikipedia’s greatest strength has always been its decentralized, community-driven model—a global village square where volunteers curate knowledge through iterative debate. This is the democracy of knowledge in action. However, this very structure contains a catch: it perpetuates biases that mirror the demographics and priorities of its most active contributors. The platform’s volunteer base is overwhelmingly male, urban, and educated, a demographic that struggles to represent the diverse linguistic and cultural tapestry of a country like India, where the majority of the population communicates in regional languages.
This disparity creates stark imbalances. The Malayalam Wikipedia, for instance, thrives with rich, local expertise, yet the English edition’s coverage of many Indian topics remains underdeveloped compared to exhaustive entries on niche Western pop culture like Pokémon. The platform’s commitment to transparency, which allows for public scrutiny via page histories and talk pages, also creates bottlenecks. A new entry on an Indian cyclone might sit unverified for days, tagged with “orange” uncertainty as editors quarrel over casualty numbers. This deliberative process is democratic but slow—a luxury in a world where real-time information reigns supreme. When Wikipedia volunteers meticulously cited sources to label Delhi’s 2020 riots, they were accused of partisan framing. Yet, the platform’s transparency at least allowed for a public debate over those sources, a messy but essential feature that its new rivals seek to eliminate.
Into this landscape of slow, human-curated knowledge steps the AI oracle. Elon Musk’s Grok promises to solve Wikipedia’s shortcomings by replacing human editors with an artificial intelligence trained on the vast expanse of the open web. The pitch is seductive: an end to human bias, instantaneous answers, and ultimate neutrality. Yet, this “neutral” algorithm is anything but impartial. Trained on a digital buffet that includes everything from newspaper archives and Reddit threads to Wikipedia itself, the model inevitably inherits and amplifies the biases of its sources.
For a country like India, where historical narratives are fiercely contested, this presents a profound danger. A question like whether the 1857 uprising was a “mutiny” or a “war of independence” is not a factual query but a deep-seated ideological debate. An AI like Grok, designed to find a single, confident answer, risks becoming hegemonic, echoing dominant Western headlines or the most frequently repeated online narratives while erasing nuanced local perspectives.
The opacity of this new model compounds the problem. Unlike Wikipedia, which allows users to trace edits and debate sources, Grok delivers polished, authoritative paragraphs with no footnotes or diff histories. It presents its conclusions as final, erasing the traces of its own decision-making process. A student in Indore copying a Grok-generated line into her project reinforces the very biases she was meant to critique, with no recourse to challenge the algorithm’s silent authority. The illusion of neutrality becomes a new form of algorithmic tyranny.
India, with its dizzying complexity, serves as the perfect microcosm of this global epistemic struggle. Its 22 scheduled languages and hundreds of dialects, its contested histories, and its deep cultural fissures highlight how both the old and new encyclopaedic systems fail to capture reality. Wikipedia’s reliance on a predominantly English-speaking, urban volunteer base means vast swathes of regional knowledge are never recorded. Grok’s reliance on English-centric data clusters only deepens this marginalization. A Telugu answer generated by Grok is often just an English answer wearing a dhoti—a surface-level translation devoid of cultural context because the model was never trained on reliable, local Telugu sources.
In this context, the choice between Wikipedia’s slowness and Grok’s opacity is a false one. The former is too slow and biased to keep pace with the modern world; the latter centralizes epistemic power in the hands of a few Silicon Valley companies, turning living civilizations into mere data points. Both systems ultimately fail to provide what a pluralistic society needs: a reliable, accountable, and culturally agile repository of shared knowledge.
Can we break the mirror of our biases and still keep the light of knowledge? Yes, but only if we fundamentally rethink the job description of an encyclopaedia. Its purpose is not to declare the final truth, but to make the path to truth visible. To escape the binary of Wikipedia’s flawed democracy and Grok’s opaque autocracy, we need a new framework built on three pillars: transparency, traceability, and cultural agility.
The foundation for such a system already exists in embryonic form: it is called structured provenance, the same technology that tracks whether your organic turmeric really came from a specific farm in Kasaragod. Applied to knowledge, it means every fact—every statement, date, and claim—is wrapped in a digital capsule. This capsule records its entire lifecycle: who uploaded the scan of the original manuscript, which librarian certified it, and whether a later scholar flagged a mismatch. The capsule cannot be edited without leaving a permanent, visible scar. It can be merged, split, or deprecated, but the original artifact remains intact, searchable, and timestamped.
Building this open ledger of knowledge will require money and, more painfully, patience. It demands a hybrid model of verification, blending the nuance of human expertise—retired librarians, local historians, subject-matter experts—with the efficiency of algorithmic curation. Volunteers will have to scan court files in Kannada, translate 19th-century Persian revenue records, and geotag ancient rock inscriptions. Universities will have to reward such drudgery with academic credit, not just social media metrics. Governments will have to treat archival metadata as critical infrastructure, as essential as highways or vaccine cold-chains.
The reward for this effort is a reference that no billionaire can buy and no state can monopolize. If tomorrow a political party insists that a medieval king’s birthplace moved overnight, the ledger will demand a new artifact—an inscription, a land deed, a carbon-dated coin—or reject the edit. The reader will see not one definitive answer but a family tree of answers, each with its own pedigree certificate. This is the future of shared memory: not a static product delivered from on high, but a dynamic, verifiable, and truly democratic process.
Brijesh Singh is a senior IPS officer and an author, his latest book is “The Cloud Chariot” (Penguin). Views are personal.
