Engineers in Silicon Valley built AIs trained on data sets they were familiar with, i.e. from that part of the Internet they themselves interact with.
Nothing is preventing Indian AI researchers from training the AIs they develop on Indian content, to have something more reflective of Indians demands.
Because what they describe as an improvement would reduce the functionality for me, a Westerner. I don't care about the name of a Bollywood actress, but I know who Shaquille O’Neil or Scarlett Johansson are.
I’m not sure your definition of colonialism is quite in line with general academic understanding of the term. There may not be explicit intent from individual developers, this does not mean power structures do not for all intents and purposes practically impose americanism on other cultures.
Meanwhile, in India, elsewhere on hn frontpage: https://news.ycombinator.com/item?id=43866321
There is far less content in Indian English than there is in American English. India has 1.4 billion people but only a small fraction (albeit one very well correlated with wealth, upper class status and access to the Internet) is literate in English.
On the upside, there would be very little Western bias in an AI trained exclusively on Hindi or Tamil content, quite the opposite.
All right, separate but equal? In my view, if you are billing your AI as a global frontier/SOTA of cognition (which Silicon Valley companies do, implicitly or not), then it should see things as a universal observer.
Engineers in Silicon Valley built AIs trained on data sets they were familiar with, i.e. from that part of the Internet they themselves interact with.
Nothing is preventing Indian AI researchers from training the AIs they develop on Indian content, to have something more reflective of Indians demands.
Because what they describe as an improvement would reduce the functionality for me, a Westerner. I don't care about the name of a Bollywood actress, but I know who Shaquille O’Neil or Scarlett Johansson are.