Thirty years of growth in outsourced tech services has transformed India's economy. However, AI's expansion is forcing the tech sector to evolve. The FT's Krishn Kaushik investigates whether data anno
By paying the Indian worker 300 rupees / 3 US dollar an hour to train the AI model, it will not change the life of the Indian worker and also take away a potential minimum wage job opportunity from an American worker. The only person to benefit will be the Silicon Valley CEO or shareholders.
Finally an honest report on AI and India. Here AI adoption is seen either as integration into legacy software or data annotation work. No emphasis on research, No explorations on other kinds of paths to AI. Maybe small labs here and there but no collective push. Everyone waits for a technology to mature and then they will adopt. The IT industries are a massive disappoint when it comes to research level work.
As an Australian, I admire India from afar. A beautiful country with great people and a heroic history far surpassing the last 200 years of misadventure. May India thrive into the future, building from its deepest intellectual, philosophical and cultural roots, and spread its great peacock feathers again.
One advantage I see for India is that it has maintained an integration and balance in its dealings with other countries, which can compensate for certain weaknesses in other areas (whether perceived or real). A wise business person will prefer to deal with a well rounded and trustworthy partner over one who is extremely powerful but untrustworthy.
"Computers offer a hydraulic of the mind that frees us from much of the drudgery of information processing in the same manner a bulldozer frees us from much of the drudgery of dirt processing.
When the drudgery vanishes, however, so do many of the drudges."
-Walter Wriston
The one gentleman said it best, "the game has changed", the companies who can understand this aspect of AI integration into everyday work tasks will succeed.
One aspect Mr. Wriston wrote about was turning managers into teachers. "As information becomes an even more essential ingredient of production, teaching will become one of the most important management skills."
What this film doesn't fully confront is the paradox at its centre — the textile workers in Karur and the homemakers being filmed are training the systems that will eventually displace them. That's not speculation. That's the pattern every automation wave follows.
But here's what makes physical AI harder than the optimists admit: I recently spoke with a senior neurologist and a gynaecologist about what AI can and cannot replicate. The gynaecologist made a point that stuck with me — human intelligence isn't just in the brain, it forms through the body's continuous interaction with the world, from the womb onward, across decades of unrepeatable lived experience. The neurologist added that even Neuralink, with direct access to brain signals, is only reading what it has already labelled — everything else remains dark.
Data annotation works for narrow, repeatable tasks. But generalising robot behaviour from a Karur textile floor to every fabric, every lighting condition, every workflow variation — that is an enormous unsolved problem. The gap between "we captured this worker doing this task" and "the robot can do any version of this task" is not a data volume problem. It is a fundamen
We are soo dumb... We train AI to replace us and we can not even afford that technology to serve back in future, some rich folks will use it,not an ordinary man.
Unfortunately, as some viewers have pointed out, the video mistakenly contains some street shots from Tamil Nadu when referring to Bengaluru. We apologise for the error.
We use cookies, device fingerprinting and cross-device tracking to personalise content, serve targeted advertising, and analyse how you use our site across your devices. By clicking Accept all you consent to the use of these tracking technologies. Your data may be used to build a profile of your interests and show you personalised ads on other sites.
Privacy policy
Comments (20)
Daily Junction discussion mixed with clearly attributed comments from the original video provider.
Join in — free. Comments on Daily Junction are for members, so real names stay rare and bots stay out.
One field. We email you a 6-digit code — no password needed. Your comment is kept while you do it.
Under 13? You’ll need a parent’s OK first — it takes them one click.
We ordinary people are really nothing to these rich people at the top.
One advantage I see for India is that it has maintained an integration and balance in its dealings with other countries, which can compensate for certain weaknesses in other areas (whether perceived or real). A wise business person will prefer to deal with a well rounded and trustworthy partner over one who is extremely powerful but untrustworthy.
When the drudgery vanishes, however, so do many of the drudges."
-Walter Wriston
The one gentleman said it best, "the game has changed", the companies who can understand this aspect of AI integration into everyday work tasks will succeed.
One aspect Mr. Wriston wrote about was turning managers into teachers. "As information becomes an even more essential ingredient of production, teaching will become one of the most important management skills."
But here's what makes physical AI harder than the optimists admit: I recently spoke with a senior neurologist and a gynaecologist about what AI can and cannot replicate. The gynaecologist made a point that stuck with me — human intelligence isn't just in the brain, it forms through the body's continuous interaction with the world, from the womb onward, across decades of unrepeatable lived experience. The neurologist added that even Neuralink, with direct access to brain signals, is only reading what it has already labelled — everything else remains dark.
Data annotation works for narrow, repeatable tasks. But generalising robot behaviour from a Karur textile floor to every fabric, every lighting condition, every workflow variation — that is an enormous unsolved problem. The gap between "we captured this worker doing this task" and "the robot can do any version of this task" is not a data volume problem. It is a fundamen