Google CEO Sundar Pichai admitted in a memo to employees that the company’s new artificial intelligence tool, Gemini, had caused offense among users due to problematic image results. The tool had generated images portraying America’s Founding Fathers as black, the Pope as a woman, and a Nazi-era German soldier with dark skin, leading to a temporary pause in its image creation feature.
Pichai emphasized that such performance was unacceptable and pledged to address the issues, promising structural changes, updated product guidelines, improved launch processes, evaluations, and technical recommendations.
Google clarified that Gemini was designed with a focus on creating diverse images to avoid biases seen in previous models. However, the AI service failed to account for situations where a varied response was not appropriate, and over time, it became overly cautious, refusing to answer some prompts correctly.
The controversy extended to Gemini’s text-generating capabilities, with outlandish responses drawing criticism. Elon Musk shared an example where Gemini was asked about the harm caused by libertarians compared to Stalin. The AI’s response initially garnered attention for being ambiguous but was later fixed to provide a more definitive answer about Stalin’s actions.
Acknowledging the imperfections of AI, Pichai expressed the company’s commitment to improving Gemini before its re-launch. Despite the challenges, he emphasized Google’s determination to meet high standards in the rapidly evolving industry of artificial intelligence.
Source: NPR
