Following controversy over historically inaccurate images, Google’s generative AI tool is under fire again by the company’s cofounder.
Sergey Brin, Google’s cofounder and former president of Google parent Alphabet, said Google “definitely messed up on the image generation,” and that he thinks “it was mostly due to not thorough testing.”
“[I]t definitely, for good reasons, upset a lot of people,” Brin said at San Francisco’s AGI House. He added that Google doesn’t know why Gemini “leans left in many cases,” but that it isn’t intentional, and other large language models could make similar errors.
“If you deeply test any text model out there, whether it’s ours, ChatGPT, Grok, what have you, it’ll say some pretty weird things that are out there that you know definitely feel far left, for example,” Brin said. He also said he“kind of came out of retirement just because the trajectory of AI is so exciting.”
Google CEO Sundar Pichai called Gemini’s generations, which included racially diverse Nazi-era German soldiers and medieval British kings “completely unacceptable” in a memo to Google staff, adding the company “got it wrong.”
Google announced it was pausing Gemini’s ability to generate images of people after some pictures stirred up social media backlash, including the anti-woke crowd. Brin noted that the backlash against AI, which faces a wider issue when it comes to bias, can also be political.
“There’s a lot of complicated political issues in terms of what different people consider misinformation versus not,” Brin said.
Pichai said in his note Google was “working around the clock” to address the issues, and is “already seeing a substantial improvement on a wide range of prompts.” Brin also said progress was being made to avoid similar issues in the future.
“If you try it starting over this last week it should be at least 80% better, of the test cases that we’ve covered,” Brin said.