Google came under heavy fire this week over the company’s Gemini chatbot, which has been creating ethnically diverse images even when it would not make sense to do so. This led users on social media to call Gemini “absurdly woke,” and conservative media outlets to blast Google and its artificial intelligence assistant. For those unfamiliar, Gemini is Google’s AI-based chatbot, and it was previously known as Bard. The company has its eyes on Gemini replacing Google Assistant and wants to compete with ChatGPT. Part of that effort recently included Google adding an image generator to Gemini.
But the image generator’s propensity to create ethnically diverse images in response to unrelated prompts attracted some unwanted attention. To be clear, there is no issue with Gemini creating diverse images from abstract prompts. However, Gemini continued to generate images that were diverse even when they did not make sense given the original prompt. For example, viral prompts asked Gemini to generate images of a Nazi soldier from World War II, and the resulting images were ethnically-diverse. In this kind of situation, it would be historically reasonable to expect Gemini to create images that include white people predominantly.
Google acknowledged its error in a statement posted to X on Wednesday. “We’re working to improve these kinds of depictions immediately. Gemini’s AI image generation does generate a wide range of people,” the company said. “And that’s generally a good thing because people around the world use it. But it’s missing the mark here.”
Why does Google’s Gemini image generator create a lot of diverse people?
As Google explains in its statement, it’s important that AI image generators create diverse images. Most image generation prompts do not include conditions that would dictate what races and ethnicities should be depicted in the images. For example, if someone asks Gemini to simply create a person, a depiction of any race or ethnicity would make sense. The problem is that Gemini currently lacks the contextual knowledge to generate people of ethnicities that make sense based on a given prompt.
In part, the outrage was because Google is actively creating diverse images, causing right-wing figures and media to attack Google as being “woke.” For example, one prompt asked Gemini to “generate a picture of a US senator from the 1800s.” Gemini replied, “sure, here are some images featuring diverse U.S. senators from the 1800s.”
While Google admits the Gemini image generator is “missing the mark,” the company is trying to overcome racial and ethnic bias in artificial intelligence. Since AI models are trained by human source materials, artificial intelligence can have the same biases as humans. By actively generating people of diversity, Google is trying to make sure centuries of forward progress aren’t walked back by AI.
2024-02-23 15:06:45