Google DeepMind CEO Demis Hassabis has addressed the recent controversy surrounding the company’s artificial intelligence image generator, Gemini, indicating plans to relaunch the tool within the next “couple of weeks.” The decision to pause access to the image generator came after users discovered that it was producing historically inaccurate images, including racially diverse depictions of figures such as the US Founding Fathers and Nazis.

Explanation of the Issue

Speaking at the Mobile World Conference in Barcelona, Hassabis acknowledged the shortcomings of Gemini, stating that the tool’s intended purpose was to showcase a “universal range of possibilities” in the images it generates, reflecting the diversity of countries where Gemini is utilized. However, he admitted that this feature was implemented too broadly, resulting in inaccuracies and potentially offensive imagery.

Commitment to Resolution

Hassabis emphasized Google’s commitment to addressing the issue promptly. He reassured attendees that the company prioritizes historical accuracy and has taken the necessary steps to temporarily suspend the feature while improvements are made. Hassabis expressed optimism about reinstating the image generator in the near future, aiming for a resolution within the next “couple of weeks.”

Apology and Acknowledgment

In a blog post last week, Google’s search lead Prabhakar Raghavan issued an apology for the shortcomings of the Gemini image maker, acknowledging that some generated images were inaccurate or offensive. Raghavan expressed gratitude for user feedback and conveyed regret for the feature’s failure to meet expectations.

As Google works to rectify the issues with Gemini, the company remains committed to delivering high-quality and culturally sensitive AI technologies. The incident serves as a reminder of the challenges inherent in developing AI systems that accurately reflect diverse perspectives while upholding standards of historical accuracy and cultural sensitivity.