Sergey Brin: Google ‘Messed Up’ Gemini’s AI Image Generation

^{This article originally appeared on Business Insider.}

The Google co-founder Sergey Brin has weighed in on the company’s Gemini drama.

Speaking at San Francisco’s AGI House over the weekend, Brin said the company had “definitely messed up on the image generation.” He added that the historically inaccurate images generated by Gemini were likely due to a lack of thorough testing.

Google paused Gemini’s image-generating feature last month after users complained it was creating strange images of people of color, including pictures depicting Black Nazis. Google’s AI model also faced criticism of some of its written responses.

Brin, who stepped back from Google in 2019, was listed in a December white paper outlining Gemini’s capabilities as a “core contributor.”

Reports have described Brin and his fellow cofounder Larry Page as being more active at the company since the launch of OpenAI’s ChatGPT in 2022.

In January 2023, The New York Times first reported that Google’s parent company, Alphabet, asked the cofounders for help after issuing a “code red” in response to OpenAI’s viral chatbot.

Since then, Brin has been spotted at Google’s Silicon Valley headquarters and been directly involved with developing the company’s AI strategy.

Brin told the audience at AGI House that he “kind of came out of retirement just because the trajectory of AI is so exciting.”

Google’s been feeling the heat in the wake of the Gemini controversy.

Some critics have portrayed the drama as evidence that left-leaning bias among Big Tech employees has affected major AI models’ output.

Elon Musk has been especially vocal, frequently pointing to an example where the bot appeared unable to say whether he or Adolf Hitler was worse. When Business Insider tested the same prompt, Gemini said it was “inaccurate and grossly inappropriate” to compare Musk with Hitler.

Brin appeared to push back on some of the criticism around the text-based model. He told the audience that any text model available, including ChatGPT or Musk’s Grok, would be likely to say some “pretty weird things” that “definitely feel far-left, for example,” if put under pressure.

Brin added that Google hasn’t fully understood why Gemini leans left in some cases and said that wasn’t the company’s intention.

Representatives for Google did not immediately respond to a request for comment from Business Insider made outside normal working hours.

On February 28, Axel Springer, Business Insider’s parent company, joined 31 other media groups and filed a $2.3 billion suit against Google in Dutch court, alleging losses suffered due to the company’s advertising practices.

Axel Springer, Business Insider’s parent company, has a global deal to allow OpenAI to train its models on its media brands’ reporting.