Google Claims It Has Fixed Gemini's AI for Image Generation
.jpg "Google Claims It Has Fixed Gemini's AI for Image Generation")
Google Restores Gemini’s People Image Generation
Google is set to reintroduce Gemini’s capability to create images of people as it expands the Imagen 3 image generator across apps and languages. The rollout will begin in the next few days, starting with early access for English-language users in Gemini Advanced, Business, and Enterprise tiers, with more languages and users to follow soon.
Addressing Past Issues
Google claims to have fixed the problems that led to the temporary suspension of this feature. They’ve made several "technical improvements," utilized better evaluation datasets, established clear “product principles,” and conducted rigorous testing to ensure the AI performs as intended. According to Google, the updated Imagen 3 under Gemini won’t create realistic or identifiable images of people. It also restricts the generation of images involving minors or content that is excessively violent, gory, or sexual. While Google admits the images won’t be perfect, they’re committed to listening to user feedback for further improvement.
Quality Over Controversy
This new version of Gemini is said to be a significant improvement over its predecessor, Imagen 2. It can produce images in specific art styles and generate pictures that more closely match the prompts provided. According to Product Manager Dave Citron, it performs "favorably" compared to other image generators.
Google initially paused Gemini’s ability to generate images of people in February after the AI produced historically inaccurate images, such as depicting people of color as German WWII soldiers. CEO Sundar Pichai called these results “unacceptable,” and Senior VP Prabhakar Raghavan noted that the model failed to consider situations where diversity was unrealistic.
A Cautious Approach
It’s still uncertain if Google has fully resolved all issues with Imagen. Often, problems with generative AI models emerge as they are tested by a larger audience. However, Google’s cautious approach contrasts sharply with xAI’s Grok, backed by Elon Musk, which allows for the creation of deepfakes and other potentially misleading content. Google remains committed to a safer, more responsible method of generating images of people, especially with major features like the Pixel 9’s Pixel Studio relying on Gemini.
.jpg&description=Google Claims It Has Fixed Gemini's AI for Image Generation){kind=link}