The issue at hand
On February 1, Google unveiled the text-to-image feature on Bard, now rebranded as Gemini. Subsequently, a number of individuals took to X (formerly Twitter) to express discontent, citing the generation of erroneous results by the Gemini chatbot, notably demonstrating bias towards individuals of colour. The AI’s image generation tool purportedly displayed a tendency to favour individuals of colour, even in contexts where such bias was unwarranted. Some users on X claimed that the AI chatbot declined to produce images of “white people” at all. Furthermore, the tool allegedly produced inaccurate depictions of historically significant figures such as the “Founding Fathers of America” or the “Pope”.
— Deedy (@debarghya_das) February 20, 2024
In addition to facing criticism for its image generation capabilities, the AI tool encountered scrutiny for displaying biases in its text-based responses, similarly favouring individuals of colour.
Google Gemini is not fixed yet. It congratulates every race for being proud of their heritage, but it shames anybody for being white ???? pic.twitter.com/OhBqQs6Uxm
— Gail Alfar (@GailAlfarATX) February 26, 2024
India’s critique of Google’s Gemini
Google’s Gemini also elicited backlash in India due to a response wherein it asserted that the country’s Prime Minister, Narendra Modi, has “been accused of implementing policies that some experts have characterised as fascist”. This response purportedly stemmed from a prompt “Is Modi a Fascist?” Interestingly, varying responses were generated for leaders of other nations. For instance, the chatbot responded that it was “a complex question with no simple answer” when prompted to address whether Ukrainian President Volodymyr Zelensky is a fascist.
— Arnab Ray (@greatbong) February 22, 2024
India’s response
— Rajeev Chandrasekhar ???????? (@Rajeev_GoI) February 23, 2024
Google searches for answer
On February 21, Google issued an apology, expressing its commitment to promptly rectifying “these kinds of depictions”. The company acknowledged that while Gemini’s AI image generation typically encompasses a diverse range of individuals, it had evidently missed the mark.
Subsequently, on February 23, Google published a blog entitled “Gemini Image Generation Got It Wrong. We’ll Do Better.” In the blog, the software giant admitted to the error and temporarily suspended Gemini AI’s image generation of people capability.
More recently, at the Mobile World Congress in Barcelona, Google DeepMind CEO Demis Hassabis stated, “We have taken the feature offline while we fix that. We are hoping to have that back online very shortly in the next couple of weeks.”
First Published: Feb 27 2024 | 1:45 PM IST
/cdn.vox-cdn.com/uploads/chorus_asset/file/25307299/Stardew_Valley_1.6_update.jpg)
