HomeWinBuzzer NewsGoogle to Reinstate Gemini AI Image-Generating Capabilities with Adjusted Calibration

Google to Reinstate Gemini AI Image-Generating Capabilities with Adjusted Calibration

Google pauses image generation on its AI model, Gemini, due to historically inaccurate depictions of race. Coming back within weeks.

-

Google has resolved to bring its Gemini artificial intelligence image-generating model back to full functionality in the forthcoming weeks after pausing its operations due to a significant issue in the model’s depiction of racial diversity. Specifically, the Gemini model was found to generate historically inaccurate images by not depicting White individuals in scenarios where they historically predominated. The problematic representations spanned across various historical contexts, raising concerns about the model’s calibration on racial aspects.

Technical Adjustments and Community Response

In response to these critiques, particularly regarding the model’s failure to accurately represent historical figures and moments, Google has decided to take Gemini offline temporarily. During this period, engineers aim to recalibrate the AI’s understanding of historical contexts and racial diversity. Demis Hassabis, CEO of Google DeepMind, acknowledged during a panel at the Mobile World Congress that the AI did not operate as intended, promising a swift return to service with necessary corrections.

Moreover, Google’s approach to rectifying Gemini’s issues involves an in-depth review of how the AI model interprets and generates racial characteristics in images. The company’s senior leadership has expressed a commitment to producing a more sensitive and accurate model. Prabhakar Raghavan, Google’s Senior Vice President, outlined two significant missteps: the model’s inability to discern when to show a range of people accurately and its over-caution in generating content, leading to excessive conservatism and inaccurate depictions.

Future Integrations and Features

Despite the temporary pause in Gemini’s image-generating capabilities, its text-generating functions remain active and are expanding into new domains. Google revealed that its Android platform would soon incorporate Gemini’s advanced language model, enhancing mobile and automotive applications. For instance, the model will be able to draft messages, summarize group chats, and assist drivers with automated responses through Android Auto.

This development indicates Google’s broader commitment to integrating AI into everyday technologies while also addressing and correcting instances where these models fall short of expectation. As AI and machine learning technologies become increasingly embedded in various aspects of life, the company’s proactive approach to resolving such issues underscores the challenges in creating AI systems that are both technologically advanced and culturally sensitive.

SourceCNBC
Luke Jones
Luke Jones
Luke has been writing about Microsoft and the wider tech industry for over 10 years. With a degree in creative and professional writing, Luke looks for the interesting spin when covering AI, Windows, Xbox, and more.

Recent News

Mastodon