Google Unveils Imagen AI as Tech Giants Hide Dangers Behind Cute Animals
Google Brain introduces Imagen, a new text-to-image AI that outperforms OpenAI's DALL-E 2 in generating photorealistic pictures. However, both companies deliberately showcase only wholesome images of anthropomorphic animals to avoid releasing problematic or harmful content.
Google Brain unveils Imagen, a new text-to-image artificial intelligence that produces highly realistic pictures and outperforms OpenAI's DALL-E 2 on standard quality metrics. Human judges consistently prefer Imagen's output over its competitors, sparking excitement about the potential end of the stock image industry. The AI generates stunning outdoor scenes and billboard-ready advertisements with just a few text prompts.
Despite these advanced capabilities, Google strictly limits the public release of Imagen and chooses to showcase only wholesome, cute images. The official demonstrations feature entirely safe concepts like anthropomorphic animals doing adorable things, such as a panda chef or a corgi in a sushi house. OpenAI relies on the exact same PR strategy for DALL-E 2, flooding social media with similarly harmless and charming animal pictures.
There is a specific technical and public relations reason for this heavy focus on cuteness. Combining innocent concepts forces the neural network to demonstrate complex understanding without generating harmful, biased, or disturbing content that these models inherently create when given broader prompts. By restricting the view to fuzzy animals, both Google and OpenAI successfully hide the darker, problematic side of their powerful image-making technologies.