Get yourself a paid ChatGPT account if you want to use the new image generator

OpenAI’s introduction of image generation in ChatGPT has been a real success, it seems, getting people excited and interested in AI image generation for the first time in a while. Maybe a bit too excited, as CEO Sam Altman said on X that the rollout to free users is delayed because of the increased popularity of the feature.

People are going mad about Studio Ghibli-style images

Originally, OpenAI intended to roll out the image generation feature to every ChatGPT user. But because of the exceedingly high demand, the company has delayed its availability for users on the free tier. CEO Sam Altman confirmed this on a post on X (formerly Twitter), saying that the popularity of the feature was higher than their already high expectations, and they need to delay its wider release. ?

The introduction of image generation has set trends viral, the most notable being the generation of Studio Ghibli-esque portraits. Members from various subscription models have taken to social media platforms to share their AI-created photographs, showcasing the tool’s creative potential. The trend speaks volumes about the growing interest and activity in AI-driven artistic tools. ?

Technical advances with GPT-4o

Image creation is powered by OpenAI’s GPT-4o model, which can process and generate various types of data, including text and images. GPT-4o employs an autoregressive approach to generating images, which produces images sequentially from left to right and top to bottom, unlike previous diffusion models like DALL-E. The method is designed to enhance the coherence and quality of generated images. ?

Despite the advancements, the feature has confronted content policy and bias issues. Users pointed out discrepancies in image generation from gendered prompts; the AI, for instance, would generate images of “sexy men” but declined similar requests for “sexy women” with a content policy violation note. Altman pointed out this discrepancy as a bug, emphasizing the company’s commitment to addressing such issues in a bid to foster equitable and respectful AI behavior.

The viral success of AI-generated images, specifically those mimicking signature art styles like those of Studio Ghibli, has spurred copyright and ethical questions. OpenAI has relaxed certain restrictions, allowing for generating images in specific styles and with celebrity likenesses. The policy shift raises questions about intellectual property rights and whether more detailed guidelines are necessary to balance the overlap of AI creativity and existing copyrights. ?