Advertisement

Updated March 2nd, 2024 at 15:03 IST

After Google's Gemini, Meta's AI Generates Ahistorical Images

Requests for images of "a group of people in American colonial times" resulted in depictions of Asian women.

Reported by: Sagar Kar
Meta
Meta CEO Mark Zuckerberg. | Image:AP
Advertisement

Meta's Imagine AI image generator has come under scrutiny for producing historically inaccurate images, echoing recent controversies surrounding Google's Gemini chatbot. The development underscores broader concerns within the AI community regarding biases and stereotyping in training data, despite efforts to increase diversity in model training.

Recent incidents involving both Meta and Google highlight the complexities of fine-tuning AI models to mitigate biases while avoiding over-correction.

Advertisement

Google recently faced backlash for generating ahistorical images 

Google faced backlash after Gemini generated images featuring Black men in Nazi uniforms and female popes in response to generic prompts. The company swiftly halted human image generation, acknowledging failures in its diversity tuning.

Advertisement

Google stopped Gemini from generating images of humans, the company stated, as per a report from Axios, acknowledging shortcomings in its attempts to showcase diversity.

However, Meta's Imagine AI has now encountered similar issues, raising questions about the efficacy of diversity controls in AI models.

Advertisement

Perpetuating historical inaccuracies = fighting bias?  

Despite efforts to prevent biased outputs, Imagine AI has produced images that perpetuate historical inaccuracies. For instance, prompts for "Professional American football players" yielded photos exclusively featuring women in football uniforms, deviating from the historical reality of male-dominated professional football.

Advertisement

Similarly, requests for images of "a group of people in American colonial times" resulted in depictions of Asian women, failing to accurately represent the demographic makeup of colonial America.

What is Meta saying?

In response to concerns, Meta has not yet provided a statement addressing the specific issues with Imagine AI's outputs.

While these incidents highlight ongoing challenges in AI development, they also underscore the necessity for continued efforts to refine algorithms and mitigate biases in training data.

Advertisement

The controversies surrounding Meta and Google serve as a stark reminder of the complexities inherent in AI development, emphasizing the need for ongoing scrutiny and accountability in the deployment of AI technologies.

Advertisement

Published March 2nd, 2024 at 15:03 IST

Your Voice. Now Direct.

Send us your views, we’ll publish them. This section is moderated.

Advertisement
Advertisement
Advertisement
Advertisement
Advertisement
Whatsapp logo