Topics
late
AI
Amazon
Image Credits:Bing
Apps
Biotech & Health
Climate
Image Credits:Bing
Cloud Computing
mercantilism
Crypto
endeavor
EVs
Fintech
fund raise
Gadgets
Gaming
Government & Policy
computer hardware
Layoffs
Media & Entertainment
Meta
Microsoft
Privacy
Robotics
certificate
societal
Space
Startups
TikTok
Transportation
Venture
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
picture
Partner Content
TechCrunch Brand Studio
Crunchboard
Contact Us
forwards of the holidays , Microsoftsaidit was upgrading the AI model behindBing Image Creator , the AI - power icon redaction tool built into the company ’s Bing hunt railway locomotive . Microsoft promised that the new model — the late version of OpenAI’sDALL - E 3model , codification - mention PR16 — would allow users to create range of a function “ twice as tight as before ” with “ higher calibre . ”
But it did n’t deliver . Complaints promptly flood X and Reddit .
“ The DALL - E we used to bonk is perish forever,”saidone Redditor . “ I ’m using ChatGPT now because Bing has become useless for me,”wroteanother .
The backfire was such that Microsoft say it ’ll restore the old example to Bing Image Creator until it can address the issues .
bring back the old dalle 3 ! the image quality is so much better on the old exemplar . like these images for illustration . the image generated by the fresh modelling soak up :( pic.twitter.com/BjIM8MS4ng
— ze ᡣ𐭩ྀིྀི ( @riegrowl)December 28 , 2024
“ We ’ve been able to [ reproduce ] some of the issues account , and design to turn back to [ DALL - E 3 ] PR13 until we can fix them , ” Jordi Ribas , point of hunting at Microsoft , saidin 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 % . ”
So what went untimely ?
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
It ’s difficult to equate model outputs from anecdotal reports , particularly when the prompts are n’t standardize . But many user said that PR16 tended to make figure reckon less realistic . Mayank Parmar , compose forWindows Latest , observe that PR16 - generate images lacked item and cultivation , and appear weirdly cartoonish and “ lifeless . ”
I do n’t cognise who you think you ’re pull the leg of with this . DALL - E is objectively worse than it ever was after this “ update ” and you ’re being outpaced by other companies like Google . It ’s dead night and day comparing image caliber now to just a distich month agopic.twitter.com/EdSdk7aign
— outwards ( @roccynoxy)December 19 , 2024
It ’s not the first time an image model that presumably passed internal checks was n’t well - receive publicly . Back in February , Google was forced topauseits AI chatbot Gemini ’s ability to make image of masses after user complained ofhistoricalinaccuracies .
The trip illustrate just how challenging it can be to measure fashion model improvements in the literal world . grant to Ribas , Microsoft ’s benchmarking found PR16 ’s quality to be “ a piece better on average ” compared to the premature Bing Image Creator model .
Whatever internal metric unit the company used , it seems clear that it did n’t align with most citizenry ’s druthers .
