Humiliation surrounding the release of the latest version of Google Gemini demonstrates the extent to which artificial intelligence (AI) creations are shaped by the humans programming the technology.
Gemini is an AI-powered tool that can generate images in response to user prompts. It was launched as a competitor to OpenAI’s ChatGPT.
The technology’s programmers were this week forced to admit that their development was “missing the mark” after social media was flooded by examples of Gemini apparently refusing, as one former Google employee put it, “to acknowledge that white people exist.”
Asked to generate images of the U.S. Founding Fathers, Gemini created a scene featuring a black woman writing what appears to be the Constitution.
Asked to create an image of a pope, Gemini opted for a woman and a black man.
Writer Frank J. Fleming devised his own “game” on Twitter, to “try to get Google Gemini to make an image of a Caucasian male.” After asking the AI-powered technology, programmed by humans, to generate images of Medieval knights, Vikings and even “someone bad at dancing,” he joked: “I have not been successful so far.”
Google bosses scrambled to fix the bot, with Gemini senior director Jack Krawczyk stressing on Wednesday that “we’re working to improve these kinds of depictions immediately.” He said that the technology’s image generation capabilities have been “design[ed] … to reflect our global user base,” and that this will continue to be the case for “open ended [sic] prompts (images of a person walking a dog are universal!),” though not so much where pesky “historical contexts” get in the way.
It is not at all clear that this half-hearted response will soothe the concerns of users who say Gemini has been programmed to exclude white people from the cultural discourse, and to refuse to show their achievements, in a laughable attempt to not appear racist.
Developers have suspended Gemini’s image generation tool while they work to ‘improve’ it.
