Google Suspends Gemini's People Generation Feature Following Multiple 'Woke' Inaccuracies

Following a controversy this week over Google’s new AI chatbot, Gemini, which generated historically inaccurate imagery—such as depicting the U.S. Founding Fathers as Black and Google’s founders as Asian—the company announced it is temporarily suspending the AI's ability to generate images of people.

Google stated it is addressing user concerns and plans to reinstate the feature in the future.

Screenshots shared by tech leaders, writers, and Gemini users reveal that the AI often creates images of people from diverse backgrounds, regardless of historical accuracy. Examples include generating Black Catholic popes (despite there never being one), portraying Vikings with dark skin (which most did not have), and depicting Nazi German soldiers as Asian (most were not).

While image-generating AI models aim to eliminate bias and promote diversity, Gemini's default outputs have sparked debate over Google's approach to representation, leading to widespread mockery and criticism.

In response to the backlash, Google’s Senior Director of Product, Jack Krawczyk, confirmed the issue and noted that the company is working to refine the model to better respect historical context while maintaining a commitment to diversity in broader prompts, such as depicting a person walking a dog. “As part of our AI principles, we design our image generation capabilities to reflect our global user base, and we take representation and bias seriously,” Krawczyk stated.

The situation prompted significant criticism from users and influential figures, including venture capitalists Marc Andreessen and Paul Graham, igniting a broader discussion about the reliability of AI chatbots like Gemini in balancing diversity and factual accuracy. This controversy has even gained mainstream attention, making headlines in The New York Post.

While Google works on a resolution, it’s worth noting that this isn’t the first instance of Gemini facing scrutiny. Since the launch of the Gemini model family in December 2023, which was promoted as comparable to, and at times exceeding, OpenAI's GPT-4, Google has encountered ongoing challenges.

The initial launch faced backlash for a staged video that exaggerated Gemini's capabilities as a digital assistant. Furthermore, a review by independent researchers found that Gemini underperformed compared to OpenAI’s older model, GPT-3.5.

This month, Google rebranded its Bard chatbot as Gemini and introduced advanced versions for subscribers, while also promoting a revamped Duet AI under the Gemini name for Workspace. Despite promising enhanced features with Gemini 1.5 and Advanced—particularly in video analysis and document parsing—these strengths may be overshadowed by ongoing controversies.

Krawczyk emphasized that reporting issues like inaccurate image generation is crucial for improvement. However, if problems continue to arise, it may undermine Google’s efforts to position Gemini as a leader in the competitive generative AI space.