When Google released its AI chatbot Bard in 2023, things didn’t really turn out to be the way the tech giant wanted. After a rather chaotic presentation (remember the presenter who didn’t have access to a demo phone?), Bard was called out for being inaccurate in one of its demo videos. Time went by and Bard, after a lot of work, started giving better responses with time. Just recently, Bard was turned into Gemini and was infused with a number of new features – image generation being one of them.
However, Gemini made headlines when people started accusing it of being “racist.” As a result, Google suspended the AI tool’s ability to generate images of individuals. And now, Google DeepMind boss Demis Hassabis says that Gemini will be back in the next “couple of weeks”.
Google DeepMind CEO on Gemini’s return
Hassabis, as reported by Reuters, said that with Gemini, things didn’t really turn out the way they were intended and that they are working on the AI tool’s shortcomings. He said that in a few weeks, Gemini’s suspended feature will be back.
“We care, of course, about historical accuracy, and so we’ve taken that feature offline while we fix that, and we hope to have that back online in very short order, the next couple of weeks, a few weeks,” he said.
The tool was not “working the way we intended”, he added.
As of now, if you try to generate an image of a person in Gemini, you will get the following response, “We are working to improve Gemini’s ability to generate images of people. We expect this feature to return soon and will notify you in release updates when it does.”
Prabhakar Raghavan on Gemini missing the mark
In a recent blog post, Google’s search lead Prabhakar Raghavan acknowledged that the Gemini image maker “missed the mark,” generating inaccurate and offensive images. Raghavan expressed gratitude for user feedback and apologised for the feature’s shortcomings.
According to Raghavan, two critical issues arose with Gemini’s human image creation feature. He explained that the tuning to ensure a diverse display of people failed to exclude cases where such variety was inappropriate. Additionally, over time, the model became excessively cautious, refusing to respond to certain prompts incorrectly by deeming innocuous queries as sensitive.
What happened with Google Gemini
To reiterate, Gemini faced criticism for generating historically inaccurate and biased images. The AI system was accused of being “too woke” and getting details about “white people” wrong. Numerous user complaints talked about Gemini’s inability to generate images of “white people” accurately. Users reported that when they requested images of figures like the pope, English kings, Vikings, or even Nazi soldiers, the AI produced pictures of individuals with darker skin tones.
In response to this, Google paused the image generation feature of Gemini and issued a statement saying that they are working on the reported issues. “While we do this, we’re going to pause the image generation of people and will re-release an improved version soon,” the company noted.
Source: https://www.indiatoday.in/technology/news/story/google-deepmind-ceo-promises-gemini-will-be-back-in-a-few-weeks-2507478-2024-02-27