Due to quality complaints, Microsoft announced that it would return its Bing AI image generator to its previous state. The company made the decision after users complained that Bing Image Creator got noticeably worse after a December update. Microsoft’s Head of Search, Jordi Ribas, said the company was able to resolve some of the reported issues. Ribas urged users to revert to PR13, Bing AI image generator’s previous model, until the issues are fixed. Microsoft recalls its Bing AI image generator upgrade The computer software company announced that it will resume the previous AI model of Bing Image Creator, the AI-powered image editing tool built into the Microsoft Bing search engine. The tech company had promised that its upgraded model, the latest version of OpenAI’s DALL-E3 model (PR16), would enable users to generate images “twice as fast as before” with “higher quality.” After Ribas posted about the upgrade in December, some users complained that Bing Image Creator was producing less-detailed results and images that didn’t accurately reflect their prompts. Complaints quickly flooded X, Reddit, and OpenAI’s community forums after the model failed to deliver. One user on Reddit revealed that the DALL-E he used to love was gone forever. Another Redditor posted that he preferred using ChatGPT because Bing had become useless to him. Jordi Ribas, Microsoft’s Head of Search, said that the company could repro some of the issues reported and planned to revert to PR13 until the issues were fixed. Ribas added that all pro users and about 25% of the requests using boosts will be on PR13. He also confirmed that the deployment process is very slow and will take the company 2-3 weeks to get to 100%. Ribas had previously mentioned that Microsoft ’s benchmarking found PR10’s quality to be “a bit better on average” compared to the previous Bing Image Creator model. Other AI image-generating software have also received backlash for not passing internal checks on image models. After users criticized its historical inaccuracies, Google had to pause its AI chatbot Gemini’s ability to create images of people. Quality complaints flood Microsoft about its upgraded Bing AI image generator One OpenAI forum member claimed he disliked how Bing placed Starburst and the lighting effects on its images. He also argued that there were suspicions the upgraded model had been trained at a lower resolution, reminiscent of upscaling and sharpening processes. The community member added that anime styles appeared more like a “coloring book” made with thick lines rather than high-resolution artwork. He also highlighted that people aiming for various artistic styles often described how VRAM, resolution, or the amount of comprehensive text had been reduced. Another user claimed that DALL-E’s recent updates destroyed its feature of implementing any image style by providing blurred updates in the foreground. One particular OpenAI user claimed that the NLP model had problems, and although it might have been improved, the training data might have been weakened. He argued that users of the upgraded model might have received reduced computing power due to the high number of requests. The chatroom member also maintained that the upgraded system needed a filter before training, and the data needed to be retrained. Some highlighted that they noticed improvements in specific aspects, such as clothing details, but the rendering problems still persisted. They hoped that the team behind the upgraded model would address the issues and revert to an earlier, more stable version of the model. Another X user urged the company to revert to the older model because the newer one was categorically worse at creating realistic-looking humans or mimicking artistic styles and photographic techniques. The company acknowledged the blowback and said that it would restore the previous model to Bing Image Creator until it can address the issues. Land a High-Paying Web3 Job in 90 Days: The Ultimate Roadmap
