Artificial intelligence
Nicole Junkermann on Artificial Intelligence in India
An enormous country, many languages at once and a generation of engineers who prefer concrete problems. Nicole Junkermann looks at India with curiosity and no forecasts.
Some countries make more sense once the figures are set aside and you look instead at what people actually do when they get up. India is one of them, which is why its relationship with artificial intelligence is worth approaching without the hurry of the headlines, which nearly always measure the thing in quantities and almost never in habits.
What follows is neither a forecast nor a recommendation. It is a note from someone who reads, travels and keeps a notebook, and who finds more than enough here for one October afternoon.
A scale that changes the questions
Size asserts itself first. When a tool reaches that many hands at once, the questions stop being abstract. Nobody is arguing about whether something could work; the argument is whether it works on an ordinary Tuesday, with patchy signal and a phone that is four years old.
That pressure has an obvious virtue: it forces simplicity. What survives in a country like that tends to be whatever genuinely settles something, and the rest is dropped without ceremony. It is a severe and fairly honest filter, of a kind you rarely manage to build in an office.
Many languages at the same time
The second trait is linguistic variety. Working there means accepting from the first day that a single language will not do, and that the same service has to be understood in several, with accents, blends and different ways of saying the same thing.
That starting point is demanding and fertile at once. Anyone who learns to build for many languages learns something that travels, including to this shore of the Mediterranean. The Leer section keeps returning to the same idea from the side of books: translating is not an extra, it is part of the work.
The useful before the spectacular
The third trait is the hardest to measure and perhaps the most valuable. In the Indian conversation about this technology, the practical carries real weight: saving a journey, avoiding a queue, explaining a form, settling a piece of paperwork without taking the day off. None of that fills a front page, and all of it changes a whole day.
The impression is of a place that understood something simple early on: a tool is judged by the time it gives back, not by the noise it makes. It is a modest measure and a difficult one to cheat.
None of this reads as a race between countries, and it should not be told that way. It is more a reminder that technology turns interesting when it lands in habits, and that habits are better watched slowly. Nicole Junkermann gathers these notes at the same pace they appear in Notas, and the rest of the notebook continues in the English notes. The next entry, on reinforcement learning, looks at the same technology from the inside.