/cdn.vox-cdn.com/uploads/chorus_asset/file/24440561/AI_art_chat_H_Jeong_1.jpg?w=696&resize=696,0&ssl=1 "Microsoft is recalling the Bing AI image generator due to quality complaints")
/cdn.vox-cdn.com/uploads/chorus_asset/file/24440561/AI_art_chat_H_Jeong_1.jpg?w=1200&resize=1200,0&ssl=1&description=Microsoft+is+recalling+the+Bing+AI+image+generator+due+to+quality+complaints){kind=link}
Microsoft is rolling back model updates to the AI-powered Bing image creator, reports TechCrunch. The rollback comes after weeks of user complaints that the tool simply didn’t work well after Microsoft “upgraded” the DALL-E 3 model on December 18.
Microsoft declined to comment on its decision to roll back the changes or provide details on what might be causing the discrepancy between user expectations and results.
Today, Microsoft’s head of search, Jordi Ribas, he tweeted that they can reproduce “some reported issues” and for now they are reverting to an older version of the DALL-E model, although it may take a few weeks before it is complete.
As soon as Ribas posted about the change in December, complaints about Bing Image Creator began pouring in less detailed results or images that did not accurately reflect their prompts. In his initial responses, Ribas stated that the model’s print quality “should be slightly better on average” than before.
All of these things are subjective and I can’t claim that any of them look better than others. If anything, this seems to be an indication that Microsoft doesn’t just have to deal with bug complaints or people unhappy with feature changes – it now also has to deal with critics of AI art comparing the machine’s performance to what they think it should produce. Maybe it’s worth asking the artists on whose work the generators were trained for tips on how to manage client expectations?
