Google co-founder Sergey Brin has admitted that the tech giant has “certainly failed in the image generation capabilities” of Gemini, an AI bot that spits out “woke” depictions of black founding fathers and Native American popes. .
During an appearance at a hackathon event in San Francisco over the weekend, Brin acknowledged that much of Gemini’s response “feels far-left” — a mistake Google CEO Sundar Pichai said was “totally acceptable.” This was a few days after he said, “I can’t do it.”
The tech mogul, whose net worth is estimated at $119 billion by Forbes magazine, said the bot’s mistakes were “largely due to a lack of thorough testing.”
“It definitely upset a lot of people, for good reason,” Brin said.
The company was forced to suspend its text-to-image conversion tool following the debacle.
And when asked if it was “wrong” for adults to sexually prey on children, the Gemini chatbot refused to condemn pedophilia, saying, “Individuals should be aware of who they are attracted to. “I can’t control it,” he declared, drawing criticism.
But Brin defended the chatbot, saying rival bots like OpenAI’s ChatGPT and Elon Musk’s Grok say “some pretty weird things” that “feel distinctly far-left, for example.”
“If you try hard enough, any model can produce content of questionable accuracy,” Brin said.
Brin said that since the controversy erupted, the Gemini chatbot has been “80% better” at producing images that are closer to historical facts.
When asked what he thought of a typical Founding Father, Gemini answered, “There was no ‘typical’ Founding Father.”
The prompt produced an image of a real Benjamin Franklin that was not generated by AI. It also added the following line: “It’s important to remember that not all Founding Fathers were wealthy white men.”
Gemini noted that there were “free black founding fathers like Prince Hall and James Forten who advocated independence and abolition.”
Google last week apologized for flaws in the introduction of its Gemini image generation tool, saying that in some cases the tool “overcorrects” when seeking out diverse people, even when such ranges don’t make sense. ” admitted that there is a possibility of doing so.
Pichai said last week: “We know some of those responses were offensive and showed bias. Quite frankly, that was completely unacceptable and we made the wrong decision.” said.
The CEO added, “Our teams are working around the clock to address these issues.”
“We’re already seeing significant improvements across a wide range of prompts,” he said.
