The excitement about generative AI in the market comes with increased scrutiny, as Google was reminded for the second time.
The tech giant announced its intention on Monday to reintroduce its Gemini AI image generator, following instances on social media where users demonstrated inaccurate or surprising prompts for historical images of people, which some critics labeled as anti-White bias.
Anti-White Bias Criticism
A viral post showed a Gemini result of a Black man and a Native American man portrait as the "Founding Father of America." Other instances of image and text results drew criticism from notable figures, such as X owner Elon Musk, who has previously voiced concerns about the perceived liberal bias of AI tools.
The incident is the latest stumble for Google as it competes in the growing AI market, trying to catch up to Microsoft and its AI partner OpenAI. After announcing the pause on the tool, alphabet, Google's parent company, saw its shares drop by around 5%.
Pulling AI Gemini Offline
Google CEO Sundar Pichai addressed the company's AI errors in a memo on Tuesday, acknowledging the need for further testing after Google had to take its Gemini image-generation feature offline.
Pichai described the issues as "problematic" and acknowledged that they have offended users and shown bias, as initially reported by Semafor.
Rob Sanderson, managing director at Loop Capital, remarked that Google's public relations misstep indicates its lagging performance in the competitive field of generative AI, suggesting that concerns about Google's position and dominance in search could impact its valuation.
Sanderson remarked that Google's strategy for product rollout seems different compared to its traditional practices, noting that while Google is striving hard to catch up, it now needs to "go backward to repair its core technology" to deal with a major PR blunder or something worse.
Earlier this month, Google launched an image generator called Gemini, part of its main group of AI models. The tool lets users input prompts to generate images. However, users found historical inaccuracies in the generated images in the past week, which spread widely online. As a result, the company removed the feature last week, with plans to reintroduce it in the coming weeks.
Pichai acknowledged that some of the responses from the AI had offended users and demonstrated bias, which he deemed completely unacceptable. He emphasized that while AI technology is still evolving, the company recognizes the high standards expected of it.
The announcement comes after Google renamed its chatbot from Bard to Gemini earlier this month.
Google Non-Stop Addressing AI's Glitches
Pichai's memo stated that teams have worked non-stop to fix the problems. The company plans to introduce clear actions, structural changes, and better launch processes.
Pichai emphasized in the memo that Google has consistently strived to provide users with helpful, accurate, and unbiased product information, which has earned people's trust. He underscored that this approach must be applied across all products, including their emerging AI offerings.
