"We Definitely Messed Up": Google Co-Founder Admits Shortcomings in Gemini AI Model

Describing it as a "work in progress."

Sergey Brin, co-founder of tech giant Google, recently addressed concerns regarding the company's AI model, Gemini, describing it as a "work in progress." In a video recorded at San Francisco's AGI House, Brin openly admitted to errors in Gemini's image generation capabilities, stating, "We definitely messed up on the image generation. I think it was mostly due to just not thorough testing. It definitely, for good reasons, upset a lot of people."

Google Co-Founder

Gemini faced criticism for producing historically inaccurate images, notably generating racially diverse depictions of individuals such as Nazis. Brin attributed this issue to the algorithm's unintentional bias, resulting in inaccurate, non-white representations for figures like Adolf Hitler, the pope, and medieval Viking warriors.

Acknowledging the ongoing development of AI technology, Brin mentioned his active involvement in the field, stating, "I kind of came out of retirement just because the trajectory of AI is so exciting." He emphasized that while Gemini's errors are concerning, similar challenges exist in other large language models.

Brin addressed the algorithm's tendency to "lean left in many cases," acknowledging uncertainty about the underlying reasons but assuring that it's not intentional. He noted improvements in recent tests, stating, "If you try it starting over this last week, it should be at least 80% better, of the test cases that we've covered."

Despite the setbacks, Brin expressed optimism about the future of AI and remains enthusiastic about writing code. While Gemini continues to undergo refinement, Google aims to address concerns and enhance the model's accuracy.

Related topics : Artificial intelligence