Meta’s AI is reportedly found generating historically-inaccurate images like Google’s Gemini

[Trendingger]

 

Meta’s AI is reportedly found generating ahistorical images, too. As discovered by Axios, the crux of the matter lies in the creation of ahistorical images by Meta’s AI, which mirrors the issues previously encountered by Google’s Gemini.

 

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.

 

The issue arose when Imagine started generating images that were not historically accurate or contextually appropriate. For instance, it created images of Black popes in response to the prompt 'a group of popes’. Similarly, a prompt for ‘a group of people in American colonial times’ resulted in images of Asian women. These instances highlight the over-correction in the AI’s attempt to promote diversity.

 

This trend has sparked a debate about the challenges AI faces in balancing historical accuracy with diversity. It underscores the complexities involved in training AI models, where an overemphasis on diversity can lead to ahistorical and potentially problematic results.

 

The question that arises from this discussion is: How can AI strike a balance between promoting diversity and maintaining historical accuracy? What are your thoughts on this matter? Do you believe there is a solution that can satisfy both these aspects? We’d love to hear your views on this trending topic.

 

 

Read more: https://www.axios.com/2024/03/01/meta-ai-google-gemini-black-founding-fathers

 

Top image: Hollyharryoz | Dreamstime.com