Microsoft recently announced an upgrade to its AI-powered image editing tool, Bing Image Creator, for the holiday season? The anticipated improvements were tied to the latest version of OpenAI�s DALL-E 3 model, known internally as PR16, which was said to boost image creation speed and quality significantly?
However, the updates quickly drew criticism from users on various platforms like X and Reddit? Many users expressed dissatisfaction with the new model, noting a decline in image quality? ‘The DALL-E we used to love is gone forever,’ commented one Redditor? Others have turned to alternatives, citing Bing�s current ineffectiveness?
Microsoft’s Response
In response to the backlash, Microsoft announced it would revert to the previous model, DALL-E 3 PR13, to address these issues? Jordi Ribas, the head of search at Microsoft, confirmed the rollback in a recent post on X? He explained that while they managed to recreate some user-reported issues, the deployment of the old model would take 2-3 weeks to reach full implementation?
Analyzing the New Model’s Flaws
The complications with PR16 don�t just stem from subjective user feedback? Technology analyst Mayank Parmar, in an article for Windows Latest, highlighted that images produced by PR16 appeared less detailed and had a lifeless, cartoonish style?
Incidents like these aren’t uncommon in the tech sphere? Earlier this year, Google’s AI chatbot Gemini faced similar issues, temporarily halting its capability to generate images of people due to historical inaccuracies?
Challenges in AI Model Implementation
These missteps underscore longstanding challenges in assessing AI model upgrades? According to Ribas, PR16�s performance should have indicated better average quality, but the metrics evidently failed to resonate with user preferences? The disconnect between internal evaluations and public reception raises questions about the effectiveness of current benchmarking practices in capturing real-world applicability?