Google DeepMind CEO, Demis Hassabis, acknowledges Gemini’s temporary removal due to backlash over historically inaccurate and racially biased image generation.
Demis Hassabis, CEO of Google DeepMind, confirms the impending return of Gemini in a few weeks, acknowledging setbacks due to unintended issues. The AI tool faced criticism for producing historically inaccurate and racially biased images, prompting temporary removal. Hassabis’ statement underscores ongoing efforts to address concerns and ensure Gemini’s improved performance upon its re-launch.
In 2023, Google’s AI chatbot Bard faced initial challenges, marked by a chaotic presentation and criticism over accuracy issues. Despite the rocky start, ongoing improvements gradually enhanced Bard’s responses over time. Recently, Bard transformed into Gemini, incorporating new features, including image generation, reflecting Google’s commitment to advancing AI technology.
Gemini, an AI tool developed by Google DeepMind, came under scrutiny and generated headlines after facing accusations of racial bias. Users accused the tool of producing racially biased images, prompting Google to take action by suspending Gemini’s ability to generate images of individuals. Now, Demis Hassabis, the CEO of Google DeepMind, has announced that Gemini will be reintroduced in the next “couple of weeks.”
Demis Hassabis, CEO of Google DeepMind, speaking to Reuters, admitted that Gemini didn’t meet expectations and acknowledged the need to address its shortcomings. He assured that the suspended feature of Gemini would be reinstated in a few weeks, indicating ongoing efforts to enhance the AI tool’s performance and mitigate concerns.
Demis Hassabis, CEO of Google DeepMind, expressed concern regarding Gemini’s historical accuracy and confirmed the suspension of that feature while improvements are made. Hassabis assured a swift resolution, stating that the feature is expected to return online within the next couple of weeks, emphasizing the company’s commitment to addressing user concerns promptly.
Presently, when attempting to generate an image of a person in Gemini, users receive the response: “We are working to improve Gemini’s ability to generate images of people. We expect this feature to return soon and will notify you in release updates when it does.”
Prabhakar Raghavan, Google’s search lead, recently addressed concerns regarding Gemini’s image generation in a blog post, admitting that the tool “missed the mark” by producing inaccurate and offensive images. Raghavan expressed appreciation for user feedback and apologized for the feature’s deficiencies.
According to Raghavan, two major issues contributed to problems with Gemini’s human image creation. Firstly, the tuning process intended to ensure a diverse representation of people failed to prevent inappropriate cases of diversity. Additionally, over time, the model became overly cautious, wrongly categorizing benign queries as sensitive and refusing to respond.
Gemini, Google’s AI system, came under fire for producing historically inaccurate and biased images, with users criticizing its portrayal of “white people” and other historical figures. Complaints surfaced regarding the tool’s inability to generate accurate depictions, resulting in images of individuals with darker skin tones being generated for figures such as the pope, English kings, Vikings, and Nazi soldiers.
In response to the backlash, Google temporarily suspended Gemini’s image generation feature and announced plans to address the reported issues. The company stated, “While we do this, we’re going to pause the image generation of people and will re-release an improved version soon,” indicating efforts to rectify the inaccuracies and biases in the AI system.
Read more Tech News
