At AI Action Summit, Prime Minister Narendra Modi stated, “If you ask the AI app to draw an image of someone writing with their left hand, the app will most likely draw someone writing with the right hand.” I tested it on Grok and ChatGPT. In both cases, AI drew person writing with right hand instead of left. I asked ChatGPT to correct it but it failed. Have a look!

PM @narendramodi today gave an excellent speech at the AI summit in France. He started his speech with an "experiment" where you ask the AI to draw a person writing with his left hand. But the AI will draw a person writing with right hand most of the time, he claimed, because it has been trained on that kind of data the most. The AI's response is biased because the training data is predominantly biased. PM Modi pointed this out as an issue in AI, to be solved using open and unbiased training data. I was surprised at how much on top of AI he really was. You should listen to his short speech. It touched all the main issues in AI and gave recommendations. I tried the experiment he suggested with Grok. And like he said, it always draws a person writing with the right hand due to its inherent training bias (as most images it must have been trained on would be right handers.) I even tried asking it to draw Obama who's a leftie. I thought, may be, it would have been trained on some images of Obama writing with his left hand and can replicate. But the training bias is so much that it drew Obama too using his right hand. PM Modi seems to have revealed AI bias to the world using such a simple experiment at the AI summit. Even take health analysis: Because the AI models are trained more on western, european health data, they may be wrongly diagnosing health issues for other populations. Or give wrong recommendations. This is why India needs its own AI, trained on uniquely Indian data - diverse data from all over India to help reduce such biases.