Microsoft Rolls Back Its Bing Image Creator Model After Users Complain of Degraded Quality 14
Microsoft temporarily rolled back its Bing Image Creator upgrade from OpenAI's DALL-E 3 PR16 to the previous PR13 version after users reported degraded image quality, including cartoonish and "lifeless" results. TechCrunch reports: Ahead of the holidays, Microsoft said it was upgrading the AI model behind Bing Image Creator, the AI-powered image editing tool built into the company's Bing search engine. Microsoft promised that the new model -- the latest version of OpenAI's DALL-E 3 model, code-named PR16 -- would allow users to create images "twice as fast as before" with "higher quality." But it didn't deliver. Complaints quickly flooded X and Reddit.
"The DALL-E we used to love is gone forever," said one Redditor. "I'm using ChatGPT now because Bing has become useless for me," wrote another. The blowback was such that Microsoft said it'll restore the previous model to Bing Image Creator until it can address the issues. "We've been able to [reproduce] some of the issues reported, and plan to revert to [DALL-E 3] PR13 until we can fix them," Jordi Ribas, head of search at Microsoft, said in a post on X Tuesday evening. "The deployment process is very slow unfortunately. It started over a week ago and will take 2-3 more weeks to get to 100%."
"The DALL-E we used to love is gone forever," said one Redditor. "I'm using ChatGPT now because Bing has become useless for me," wrote another. The blowback was such that Microsoft said it'll restore the previous model to Bing Image Creator until it can address the issues. "We've been able to [reproduce] some of the issues reported, and plan to revert to [DALL-E 3] PR13 until we can fix them," Jordi Ribas, head of search at Microsoft, said in a post on X Tuesday evening. "The deployment process is very slow unfortunately. It started over a week ago and will take 2-3 more weeks to get to 100%."
Typical user complaint: (Score:2)
all the dicks and boobs are pixelated!
"degraded quality" indeed. ;)
Re: (Score:2)
Completely unfappable.
Once again, no testing (Score:2)
Just like their updates for Windows, it is clear Microsoft does no testing before pushing software out to the masses. Even a cursory attempt at creating images would have shown these issues. Instead, in an effort to save a few bucks, they let the end user be the tester. They'd rather look foolish and incompetent than have to explain how they didn't make their quarterly numbers because they spent money on testing their software.
Re:Once again, no testing (Score:4, Informative)
Yep. Their stuff is getting slowly more and more crappy. The next large IT incident will be caused by sheer Microsoft greed and incompetence.
Re: (Score:2)
Just like the last one and the one before that?
Re: (Score:2)
It is like that
"I don't always test my code but when I do, I do it in production" meme that was around a lot at some point.
Like the one on: http://pbs.twimg.com/media/BY5... [twimg.com]
Two to three weeks for deployment? (Score:1)
How can it take two to three weeks (more) to deploy the downgrade? Have they interconnected their application servers using 300bps modems or something?
Re: (Score:3)
In the real world where you run many servers in parallel serving many users and have a requirement to test things carefully to prevent causing a global outage rolling something back is more complicated than hitting the "undo" button.
Re: (Score:2)
Readers added context
There are virtually no instances of Microsoft incompetence that thegarbz won't fend for.
Interesting, it's like you don't read my posts at all. There's plenty I don't fend Microsoft for. Just look at literally any article about a botched windows update rollout.
But in any case the readers adding the context are begging the question. What is "incompetent" about taking your time to do a job properly and not rushing to screw the system?
Re: (Score:2)
In the real world where you run many servers in parallel serving many users and have a requirement to test things carefully to prevent causing a global outage ...
... you have a rollback process ready to go before you ever begin the rollout! How TF do you trick your brain to jump from that premise, one that focuses on proper enterprise level planning, into justifying a weeks long rollback???
Had you said, "In the real world where you're understaffed and overworked and forced to managed many servers with little support, you don't have time to test things carefully and ensure your rollback process is ready to go," then you may have had a point. There's obviously a reaso
Re: (Score:2)
They probably do. But again rolling out and rolling back is not a case of clicking an undo button. You're making an assumption about how long it took them to rollout in the first place.
Re: (Score:2)
But again rolling out and rolling back is not a case of clicking an undo button.
Rolling back certainly could be as easy as a click of a button. That takes more planning and work up front than to punt on the rollback process and hope you don't need it. The fact it will take weeks to roll back means they did NOT do the necessary planning to make it as easy as a button click to roll back.
Feel free to argue that such rollback planning is unwarranted or overkill or whatever excuse you want to use, but better planning would have minimized the rollback timeframe. Don't be daft.
Re: (Score:2)
Probably took them a while to realize that no, their users were _really_ not willing to accept this crap.