Meta Inaccurate AI Images Followed by Google Gemini: An AI Bluff

Facebook’s Parent Meta is facing the issue of historically inaccurate image generation with its tool Imagine AI (Artificial Intelligence) image generator. Earlier, Google Gemini Chatbot and now Meta are facing a lot of criticism and controversies surrounding their AI image generation tool.

Despite efforts being put in to enhance diversity in model training, the AI community fails to remove stereotypes and gender biases from data. These recent events have raised concerns for both Meta and Google and the need to fine-tune their AI models in order to minimize biases.

In a recent event, Google Gemini produced images of black men in Nazi uniforms and female popes when given generic prompts. Google, when receiving criticism for the same, promptly stopped human image generation, accepting the fact that its AI model needed diversity tuning efforts.

Meta’s AI, known as Imagine, is based on the Emu image-synthesis model. This tool, integrated within Instagram and Facebook DMs, generates images from text prompts. However, it has been observed that Imagine is producing images that are historically inaccurate, similar to the gaffes made by Google’s Gemini. 

Also Read: Gemini AI Blunder Costs Google $70 Billion Loss: Pischai Calls It Unacceptable and Warned Employee

As per the reports, Meta’ Imagine AI also generated similar outputs, which raised anger, leading to questions about the accountability of these artificial intelligence systems. For example, when the prompt “Professional American football players” was given, Imagine AI gave an output solely of women in football uniforms. This resulted in questioning the historical reality of the male-dominated professional football system. 

Another example is when Imagine AI was requested for images of “a group of people in American colonial times” – It resulted in the output of Asian women. This showed that the AI system failed to understand the correct demographic composition of colonial America.

As discovered by Axios for the creation of ahistorical images by Meta’s AI: 

Also Read: Gemini Chatbot rolled out by Google for Android Users

Even though Meta has not officially released any statement regarding the historical image generation by its Imagine AI tool, these systems need efforts to refine algorithms and alleviate biases in training data.