recently announced that its
Gemini
team has shown significant progress in improving the company's AI
image recognition
tool. The tech giant’s head of search,
Prabhakar Raghavan
made this announcement during a recent all-hands meeting at Google's Mountain View headquarters. At this meeting, Raghavan provided an update on the company's efforts to correct its AI-generated
historical images
.
Earlier this year, Google’s ChatGPT rival was criticised for inaccurately portraying people of different races. The issue gained attention when Gemini’s image generator model started producing incorrect images that misrepresented historical figures. In these pictures, the AI chatbot replaced White individuals with people of Black, Native American and Asian descent.
In February, Google withdrew Gemini’s image-generation feature after widespread viral criticism.
What Raghavan has to say about Google’s Gemini team
According to a report by CNBC, Raghavan praised the teams working on Gemini, during the meeting. He said that the core group of employees looking after the company’s AI models have stepped up their efforts and increased their working time from 100 hours a week to 120 hours to correct Google’s image recognition tool within days. Raghavan revealed that this effort helped the team to fix nearly 80% of the issues in just 10 days.
However, Google still hasn’t reintroduced Gemini’s ability to generate images of people. Soon after Gemini’s image generation tool was taken down, Google’s AI leader, Demis Hassabis said that it would be re-released in weeks. Raghavan also clarified that the lack of effort from the company wasn’t the reason behind the
AI tool
's failure to generate accurate images.
“I want to be clear, this wasn’t some case of somebody slacking off and dropping the ball,” he added.
During his speech, Raghavan also reminded the audience about the company’s ability to respond quickly on important matters. He highlighted an effort in 2023 when the Bard team (now Gemini) and Magi team, which focuses on AI-powered search, launched products within months. Raghavan suggested that it was something the company couldn’t have accomplished with bigger numbers.
“The realisation was ‘gosh, if we had thrown 2,000 engineers at these projects, we wouldn’t have got it done,’” he said.