“How long have we left?”

Artificial intelligence not only develops, but is accelerating at a pace threatening to bypass the service -based economy in India. The new data shared by the founder of Zerodha Nithin Kamath is a blatant picture: Artificial intelligence is now able to perform longer and more complex tasks than ever, and improves very quickly so that there are not many human functions.

One of the plans reveals that the length of the tasks that Amnesty International can carry out with a success of 50 % is doubled every seven months. Just a few years ago, models such as GPT-2 seconds can be addressed. Today, GPT-4O can keep it for about an hour.

And the most advanced models-such as Sonnet 3.7-already dealt with multi-watch functions such as photo worksbooks. These are no longer experiences. They are alternatives.

The second chart is more concerned. It sets the success of Amnesty International through the difficulty of the task and the duration. In clean and short tasks, higher models now achieve more than 90 % of the accuracy. Even the most chaotic and dizzy problems – that are viewed beyond the reach of artificial intelligence – see sharp gains. Only the most difficult category, long and chaotic tasks, are still less than 30 % of success.

But even this barrier erodes.

In the face of this data, Kamath asked a great question on X: “How long have we left?”

It is not alone in raising the alarm. Atomberg Arindam Paul is estimated that 40-50 % of white collar function in India-especially in IT and BPO- can disappear.

A recent study by IM Ahmedabad showed that 68 % of white collarship professionals believe that artificial intelligence will automate their jobs, at least partially, within five years. Jefferies expects that half of all white collars for beginners-from marketing to sales to software-within 1-5 years only.

This is not theoretical. It happens. India, with its strong dependence on the services and the limited manufacturing capacity, is uniquely weak.