Skip to content

Gemini, Google's AI, to Return Generating People Images

  • Google's multimodal generative AI, Gemini, is expected to resume generating images of people soon.
  • The suspension followed concerns about Gemini producing historically inaccurate depictions of individuals.
  • Demis Hassabis acknowledged the complex challenge and hinted at future improvements in historical figure depictions.

Google is optimistic that its multimodal generative AI tool, Gemini, will soon regain the ability to generate images of people. According to Demis Hassabis, the founder of DeepMind, this capability should be back online in the next few weeks. 

Google had temporarily suspended Gemini after users noticed that it was producing historically inaccurate images, such as depicting the US founding fathers as a diverse group instead of only white men. 

Hassabis explained that the issue arose because Google failed to recognize when users were seeking a "universal depiction" rather than specific details. He acknowledged that this is a complex challenge in the field of advanced AI and hinted at future improvements to address prompts for historical figures.

Addressing the misuse of generative AI tools by authoritarian regimes, Hassabis noted that this is a complex issue that requires a broad societal response.

It involves research, debate, and collaboration with various stakeholders to set limits and ensure values align with the use of these systems. As AI technology advances, preventing misuse by bad actors becomes increasingly important.

Edited by Shruti Thapa