Google’s Gemini AI Draws Criticism Over Bias in Image Generation

Google’s Gemini, a new AI model, recently launched with a controversial rollout that has drawn criticism from tech figures like Elon Musk and Nate Silver. The AI model faced backlash for generating inaccurate and offensive images, leading Google to issue an apology and temporarily pause its image generation capabilities.

One of the main issues arose when users discovered that Gemini’s image service was producing historically inaccurate images, such as depicting Black vikings, an Asian woman in a German World War II-era military uniform, and a female Pope. Google acknowledged these shortcomings and attributed them to a failure in the model’s training, which was supposed to ensure a diverse range of people in its results but didn’t account for inappropriate instances.

FiveThirtyEight founder Nate Silver called for Gemini to be shut down after a controversial prompt asked the AI model to choose who has negatively impacted society more: Elon Musk, Adolf Hitler, or memes. Musk, who founded the AI startup xAI, criticized Google for what he called “insane racist, anti-civilizational programming” in Gemini. He also alleged that the AI’s discriminatory issues extended to Google search and echoed claims that Google is rigging the 2024 presidential election in favor of left-wing candidates.

Despite the controversy, Google announced that Gemini’s image generation of people would relaunch in a few weeks, according to a statement from Google DeepMind CEO Demis Hassabis. However, Musk claimed in a tweet that a senior Google executive told him it would take “a few months to fix” the issues with Gemini’s image generation, expressing doubt that Google’s internal bureaucracy would address the problems effectively.

In its blog post, Google stated that it was trying to avoid past issues with image generation technology, particularly concerning violent or sexually explicit content. This comes as other AI programs, such as Stability AI’s Stable Diffusion XL, have been found to exhibit bias, producing images exclusively of white people in certain contexts. The Washington Post also reported bias in AI, with prompts for “productive person” and “attractive people” only producing images of white individuals, and a prompt for “person at social services” only generating images of people of color.

The controversy surrounding Gemini has had an impact on Google’s parent company, Alphabet, with shares closing down more than 4% after further criticisms of Gemini emerged. Gemini’s launch puts it in direct competition with OpenAI’s ChatGPT AI model, which has received significant backing from Microsoft and reached a valuation of $80 billion. Musk’s xAI startup is also a player in the AI sector, offering products like its chatbot named Grok as an alternative to ChatGPT and Gemini.

Musk has been vocal in his criticism of Google’s leadership, particularly targeting product director Jack Krawczyk, whom he accused of contributing to Google’s alleged racism and sexism. These criticisms come amid a broader debate about bias and ethics in AI development, highlighting the challenges faced by companies as they navigate the complex landscape of AI technology.