Every week, I sit with my two teenage sons and we go through the hashtagBhagavadGita together, more than religious instruction for philosophical training. I want them equipped with the deepest Vedantic concepts our civilisation has produced before they fly off the nest for higher studies, because I suspect the world they are flying into will ask them “who are you” with a force and frequency I never faced at their age.
I do not know how hashtagAI is going to change everything. What I do know is what I am trying to do with whatever time and capability I have. Build my sons’ arsenal of self-knowledge deep enough that nothing can hollow out their sense of who they are. And in parallel, become more deliberately involved in questions of culture and identity myself, so that this extraordinarily powerful technology gets pointed, even in some small way, toward making the world better for them.
Here is the elephant in the room, named directly. We are at real risk of AI quietly homogenising human culture. A model trained mostly on English-language internet data does not just answer in English. It begins to think in the assumptions, the references, the aesthetic instincts of whoever produced the most text online. Without deliberate correction, “global AI” risks becoming, by sheer statistical gravity, a culturally narrow AI wearing a universal costume. A child anywhere on earth could grow up asking an AI about identity and receive an answer rootless to their own soil.
There is a second risk, say if a teenager can ask an AI about the hashtagGita instead of sitting weekly with a parent who has wrestled with it personally, something in the chain of transmission breaks, even if the algorithm’s answer is technically correct.
And underneath both risks lies the deepest one, that if AI becomes capable enough at everything, will an entire generation simply stop asking who they are at all, because the question stops feeling necessary?
I do not think the answer is rejecting the technology because the tool itself is not the problem. The question is what steers it. There are already real examples of AI pointed in the right direction. The Cherokee Nation partnered with Microsoft to build an AI-powered Cherokee keyboard and text-to-speech tools, now used in tribal schools where fewer than 2,000 fluent speakers remain. Manx, a Celtic language UNESCO had declared extinct, is being taught to children again and now available inside major translation platforms. Meta’s No Language Left Behind project built translation models for over 50 African languages. In each of these cases, the technology became a life raft for the culture to keep speaking in its own voice, to its own children, on its own terms.
My sons will leave home soon enough. I cannot control what AI becomes in the years they are away from home. What I can control is whether they leave carrying a hashtagVedantic arsenal substantial enough to ask better questions than the ones any model hands them by default.