Following a controversy that erupted earlier this week when users realized Google’s new AI chatbot Gemini was generating ahistorical and inaccurate imagery — such as making the U.S. Founding Fathers black and Google’s own founding duo Asian — the company announced today that it is temporarily suspending the AI’s ability to generate images of people entirely.
Google previously said it is working to fix the issues raised by users, and plans to reinstate the feature at a later time.
What went wrong with Gemini?
According to screenshots of interactions shared by tech leaders, writers and general users of Gemini, the chatbot generates images of people of color and different ethnicities even in cases of historical context where they are not applicable, such as generating black Catholic popes (there has never been one), giving Vikings dark skin (most didn’t have it) and making Nazi German soldiers look Asian (most were not).
Image generating AI models are often tuned to avoid bias and focus on diversity, but Gemini’s default generations of inaccurately broad diversity left many debating Google’s approach to bias and many others mocking, criticizing and trolling the company.
VB Event
The AI Impact Tour – NYC
We’ll be in New York on February 29 in partnership with Microsoft to discuss how to balance risks and rewards of AI applications. Request an invite to the exclusive event below.
Request an invite
In response to the flurry of posts flagging the problem, Google’s Senior Director of Product Jack Krawczyk took to X confirming the issue and that the company is working to fix it.
He said that the company will tune the model involved to accommodate historical nuances but will continue to focus on diversity for open-ended prompts, like a person walking a dog, to reflect global user base.
“As part of our AI principles…, we design our image generation capabilities to reflect our global user base, and we take representation and bias seriously,” Krawczyk wrote in the post.
Still, the issue attracted an avalanche of criticism from users and prominent tech leaders, including venture capitalists Marc Andreessen and Paul Graham, and raised the larger debate around if AI chatbots such as Gemini could be trusted to be truthful or were overly censorious in the name of modern standards of diversity and equality. The issue even penetrated beyond the tech industry, making the front page of New York City’s local newspaper, The New York Post, today:
Not the ‘Gemini era’ Google expected
While Google continues to work on fixing the issue, it is important to note that this is not the very first case of Gemini making headlines for all the wrong reasons.
The company unveiled the Gemini family of models after much hype in December 2023 and has since been on a roll to bring the “Gemini era” front and center. Google promoted Gemini as a leading AI model comparable to, and in some cases, surpassing OpenAI’s GPT-4 (which powers ChatGPT)
But soon after launch, it was criticized for releasing a staged video that misleadingly overstated Gemini’s capabilities serve as an intelligent digital assistant capable of sophisticated conversation and assistance with daily tasks, and a review by independent researchers found that Gemini was actually worse than OpenAI’s older LLM, GPT-3.5.
It rebranded the Bard chatbot as Gemini earlier this month, launching an advanced version for paying subscribers, and even rebranded and relaunched Duet AI as Gemini for Workspace.
The newer versions of Gemini, 1.5 and Advanced, promised to be much more powerful. Anecdotal examples of users indicate the latest version of the chatbot does have impressive capabilities, especially with video analysis and summary, as well as parsing numerous documents using its extremely long context. Yet these capabilities may be overshadowed in the current and ongoing controversy.
Krawczyk wrote that flagging problems like inaccurate image generation helps with the alignment process – iteration on feedback. However, if issues keep cropping up regularly, it will become difficult for the Sundar Pichai-led giant to make the case that Gemini is superior to other gen AI offerings.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.
