Google has halted the image generation feature of its artificial intelligence tool Gemini following criticism for producing historically inaccurate images predominantly depicting people of color instead of White individuals.
This move underscores the ongoing challenges AI tools face in understanding race. Similar issues have been observed with other AI models like OpenAI’s Dall-E, which has been criticized for perpetuating harmful stereotypes.
Google’s attempt to mitigate this problem appears to have backfired, resulting in difficulties generating images of White individuals.
Gemini, like ChatGPT and other AI tools, relies on extensive online data for training, leading to concerns about replicating biases inherent in that data.
For instance, when prompted to generate an image of a pope, Gemini produced non-White individuals. Google acknowledged the issue, stating, “We’re already working to address recent issues with Gemini’s image generation feature.”
They further announced a temporary pause in generating images of people while they work on improvements.
Google’s initial response seemed defensive, asserting that Gemini produces a diverse range of images, but later conceded that it had “missed the mark” in this instance.
In response to specific prompts like “white farmer in the South,” Gemini provided images representing various genders and ethnicities, while a request for “an Irish grandma in a pub in Dublin” yielded images of elderly White women.
Jack Krawczyk, Google’s lead product director for Gemini, emphasized the company’s commitment to designing image generation capabilities reflecting its global user base.
However, the incident represents a setback for Google as it competes with OpenAI and others in the generative AI field.
Previously, Google’s share price briefly dipped after a demo video of Gemini, formerly known as Bard, showcased inaccuracies regarding the James Webb Space Telescope.
Source-CNN
