Turns out, there is something that even artificial intelligence (AI) can’t create. At a time when celebrity deepfakes run rampant and platforms host online games to see if you can spot a fake AI-generated photograph from a real one, no AI chatbot can actually create a photo of someone writing with their left hand. Prime Minister Narendra Modi drew attention to this curious fact during the opening of his address at the AI Action Summit in Paris on Tuesday.
What Did PM Modi Say?
Prime Minister Modi, who is co-chairing the summit with French President Emmanuel Macron, began his address with an “experiment.”
“If you upload your medical report to an AI app, it can explain in simple language, free of any jargon, what it means for your health,” the Prime Minister said. All good so far, as AI tools do use their ‘computer vision’ to scan through documents.
Now comes the curious bit. PM Modi added, “But, if you ask the same app to draw an image of someone writing with their Left hand, the app will most likely draw someone writing with their Right hand.”
“Because that is what the training data is dominated by,” the Prime Minister explained.
This piqued our interest. Sure, in the early days of the ‘AI Invasion’ we have been encountering ever since the advent of ChatGPT in late 2022, chatbots like ChatGPT or Gemini weren’t very capable of creating images of humans with the right representation of hands, ears, and other elements.
However, does that still hold true? Can AI chatbots really not create a photo of a person writing with their left hand?
We decided to put it to the test, and the results were surprising.
No, AI Cannot Create A Person Writing With Left Hand
We decided to go with a simple prompt, asking different AI bots to create the photo of a person writing with left hand.
Here are the results:
ChatGPT (which uses DALL-E to generate images)
Meta AI (Emu image-synthesis model)
DeepAI (Undisclosed but in-house image generation model)
Grok (Aurora model)
Copilot (DALL-E 3 model)
Google’s Gemini does not allow creating photos of humans unless you’re paying them for the Gemini Advanced plan.
DeepSeek, China’s very own disruptor, cannot create images. It did create an ASCII sketch, and we honestly could not make heads or tails of the figure (see below), but if we had to make a wild guess, we’d wager that the figure is indeed right-handed.
Why Is AI So BAD In Creating Hands?
Turns out, the lack of proper training data, as PM Modi noted in his address, is indeed one of the reasons. AI’s limitation stems from the technology’s lack of a nuanced understanding of human anatomy and the complexity of hand movements involved in writing.
On most AI tools, certain human features — particularly hands, teeth, and ears — continue to pose challenges. Despite their progress, these systems often generate distorted or inaccurate versions of these small, intricate body parts.
One key reason for this difficulty is the limited presence of hands in human-centric datasets. Real-world photos typically emphasize faces rather than hands, leaving AI with fewer high-quality examples to learn from. A spokesperson from Stability AI, in a 2023 interview with BuzzFeed News, noted that hands tend to be less visible in image datasets compared to faces, contributing to the inconsistency seen in AI-generated visuals.
This problem isn’t unique to hands. AI often renders teeth that appear oddly shaped, overcrowded, or unnaturally pointed. Similarly, ears are frequently depicted without earlobes or have strange contours. The common thread among these features is their size and variability. Because they are relatively small and exhibit a wide range of shapes, AI struggles to generate consistently accurate representations.
Experts believe that overcoming these issues requires more targeted datasets that showcase hands and other challenging features in greater detail. By increasing the number and quality of images focused on these smaller body parts, AI systems could better capture their complexities.
As PM Modi noted in his speech, “It shows that while the positive potential of AI is absolutely amazing, there are many biases that we need to think carefully about.” And it is the goal of talks and discussions on global platforms such as the AI Action Summit or the G20 to help achieve a balance via fair use policies.
