Recent advancements in AI image generation technology have significantly improved the realism of artificial images. Tools like OpenAI’s DALL-E, introduced around five years ago, have evolved from producing small, low-resolution images to much more detailed graphics. While early iterations were identifiable by strange features, contemporary models have made strides in mimicking lifelike details.
DALL-E 2, which launched a year later, created images at higher resolutions but still exhibited recognizable inconsistencies. Other platforms, such as Midjourney and Stable Diffusion, have also introduced enhanced capabilities, focusing on reducing obvious flaws and rendering text with greater accuracy. Despite these improvements, many AI-generated images previously maintained an overly polished appearance, lacking the subtle imperfections characteristic of real photographs.
Recent developments have shifted focus towards achieving greater realism by emulating the aesthetic seen in smartphone photography. Google’s recent release of the Nano Banana model within its Gemini app emphasizes a more authentic look, resembling the style of images captured by mobile devices. This model draws from real-world knowledge to produce images that are deemed more faithful to actual likenesses compared to other AI tools.
Meanwhile, Google’s effort to implement the Content Credentials standard aims to provide transparency regarding the origins of images. This standard would enable every image taken with Pixel 10 series cameras to carry a cryptographic signature, indicating whether it was generated by AI or captured through traditional methods. New features within Google Photos are facilitating the visualization of these credentials, although for widespread effectiveness, further adoption by hardware manufacturers and social media platforms is necessary.
As AI-generated images become increasingly realistic, this raises concerns about authenticity and the ease of distinguishing between real and fake imagery, prompting ongoing discussions around the implications of such advancements in technology.
Source: https://www.theverge.com/column/843883/ai-image-generators-better-worse