Google has taken immediate steps to address and rectify reported biases within its conversational AI system, Gemini, specifically within its image generation capabilities. The tech giant has identified that Gemini disproportionately depicted people with darker skin tones when requested to generate images of individuals in both historical contexts and fictional scenarios. In response, Google announces the temporary suspension of this feature to prevent the propagation of these biases.
Commitment to Diversity and Inclusion
Acknowledging the issue, Google emphasizes that diversity and inclusion are core principles guiding its technologies. The company has voiced its intention not to allow its products to amplify real-world biases. A spokesperson for Google conveyed that efforts are underway to enhance Gemini’s image generation feature, ensuring it more accurately represents people of all ethnicities and backgrounds. While the timeline for these improvements remains unspecified, the commitment to re-launch a more equitable version is clear.
We're already working to address recent issues with Gemini's image generation feature. While we do this, we're going to pause the image generation of people and will re-release an improved version soon. https://t.co/SLxYPGoqOZ
— Google Communications (@Google_Comms) February 22, 2024
Technological and Societal Implications
Experts in artificial intelligence highlight that biases in AI systems frequently stem from the datasets used during their training phases, which often lack comprehensive representation of diverse populations. Google’s decision to refine Gemini before reintroducing its image generation capability aligns with broader industry and governmental moves toward equitable AI development. Notably, earlier this year, a consortium including Apple, OpenAI, and Microsoft was established to form protocols around the safe and bias-free development of generative AI technologies.
Google’s initiative to address and correct the biases in Gemini’s AI points to a growing acknowledgment within the tech industry of the responsibility to develop AI in a manner that is not only technologically advanced but also ethically sound and inclusive. As the company works on improvements, the potential for AI to reflect and respect global diversity remains a critical focal point.
Gemini is also facing issues with accurate historic information. For example, users highlighted instances where Gemini depicted a Native American man and an Indian woman as an 1820s-era German couple, an African American Founding Father, and Asian and indigenous soldiers as members of the 1929 German military. Jack Krawczyk, Google’s Senior Director of Product overseeing Gemini, acknowledges the inaccuracies and assures users that the team is actively working to rectify these issues.