Artificial Intelligence and the Integrity of Sanskrit Knowledge

The educational value of a digital answer depends on whether a reader can examine the material supporting it. This is especially important when a system interprets a Sanskrit passage whose meaning may depend on grammar, context and a history of commentary. A fluent explanation can assist inquiry only when its relationship to the source is clear.

My proposal for a Centre for Sanskrit Language Models and Knowledge Systems begins with that responsibility. Scholars and engineers would work together on the material from which tools are developed and on the standards by which their outputs are assessed. The programme would serve research and learning, with scholarly judgement guiding its design.

The first requirement would be a well-described collection of sources. Editions, commentaries, translations and recordings should retain their attribution and context. Different versions should be distinguishable. The system should allow a researcher to identify the passage being consulted and understand the status of the material. These arrangements would make it possible to investigate an answer rather than simply accept it.

Language tools could then support tasks such as searching, grammatical analysis and comparison. Their performance would need testing by specialists. An error in identifying a grammatical relationship may alter an interpretation, while a translation may conceal an important ambiguity. Evaluation should include difficult cases and should examine how the tool communicates uncertainty when reliable information is unavailable.

I would also place limits around generated explanations. A system should distinguish a source passage from its own summary, identify the commentary it follows and acknowledge substantial interpretive disagreement. It should not present an unsourced statement as the settled meaning of a tradition. Teachers would help students learn when a tool is useful and when direct consultation of texts and specialists is necessary.

The programme could begin with a defined collection and a narrow set of tasks. Results should be documented so that scholars can assess what has worked and what requires revision. A carefully evaluated resource would provide a sounder basis for expansion than an ambitious promise covering every field at once.

My interest in artificial intelligence arises from the possibility of extending access to demanding knowledge. That possibility deserves serious work. Its educational contribution will depend on source quality, transparent interpretation and continuing scholarly oversight, with technology helping readers become more capable investigators of the traditions they study.