Captionbot is another AI from Microsoft, and it's not doing so hot either

By Lulu Chang Published April 17, 2016

microsoft captionbot screen shot 2016 04 17 at 3 26 32 pm — Image used with permission by copyright holder

You’d think that after its last AI chat bot fiasco, Microsoft would’ve learned its lesson — or at least waited a bit longer before introducing another monster to the world. Luckily for us, history has a tendency to repeat itself, and the company’s latest creation is a sometimes hilarious entity known as Captionbot. The premise is simple — recognize the contents of a photo, and come up with an appropriate caption. But alas, while the concept may be straightforward, the execution isn’t always so cut and dry. And social media has been quick to point out all of Captionbot’s faux pas.

In one case, the artificial intelligence system thought that Michelle Obama was a cellphone. And there’s also the time when the dress that launched a thousand debates (you know, the blue and black or white and gold dress) confuses the bot as much as it confused all of us. The system thought the dress was a cat wearing a tie (or in some cases, a suitcase).

Recommended Videos

The bot also looks to be attempting to rewrite history, absolutely refusing to identify Hitler as the infamous Nazi leader he was. When CNNMoney fed the AI photos of the dictator and swastika, more often than not, Captionbot responded, “I really can’t describe the picture” alongside a confused emoji. Strangely enough, it could identify other famous Nazis like Joseph Mengele and Joseph Goebbels.

Please enable Javascript to view this content

Osama bin Laden is another infamous face Captionbot can’t recognize, although it is unclear whether this is due to restrictions Microsoft put in place to prevent Tay-esque disasters from taking place. “We have implemented some basic filtering in an effort to prevent some abuse scenarios,” a spokesperson noted of the limitations.

You can upload your own photos for Captionbot to analyze, but proceed at your own risk. Sometimes, this bot gets it very, very wrong. Just check out a couple of our favorite examples below.

#captionbot thinks Michelle Obama is a cell phone. pic.twitter.com/FtC6tP6Ben

— David Sim (@davidsim) April 14, 2016

Captionbot.ai 's view of my avatar: pic.twitter.com/OUctTxhx0p

— Richard Gadsden (@po8crg) April 14, 2016

https://twitter.com/ralphharrington/status/720544025649025024?ref_src=twsrc%5Etfw

Topics

Former Digital Trends Contributor

Fascinated by the effects of technology on human interaction, Lulu believes that if her parents can use your new app…

Computing

OpenAI spills tea on Musk as Meta seeks block on for-profit dreams

OpenAI has been on a “Shipmas” product launch spree, launching its highly-awaited Sora video generator and onboarding millions of Apple ecosystem members with the Siri-ChatGPT integration. The company has also expanded its subscription portfolio as it races toward a for-profit status, which is reportedly a hot topic of debate internally.

Not everyone is happy with the AI behemoth abandoning its nonprofit roots, including one of its founding fathers and now rival, Elon Musk. The xAI chief filed a lawsuit against OpenAI earlier this year and has also been consistently taking potshots at the company.

Computing

OpenAI’s Sora doesn’t feel like the game-changer it was supposed to be

Sora's interpretation of gymnastics

OpenAI has teased, and repeatedly delayed, the release of Sora for nearly a year. On Tuesday, the company finally unveiled a fully functional version of the new video-generation model destined for public use and, despite the initial buzz, more and more early users of the release don't seem overly impressed. And neither am I.

https://x.com/OpenAI/status/1758192957386342435

Computing

A new test shows Microsoft Recall’s continued security problems

Recall screenshot.

Microsoft is currently previewing its latest version of Recall to Windows Insiders on Snapdragon-, Intel-, and AMD-based Copilot+ PCs -- and the topic on most users' minds is security. The company updated its security and privacy architecture for the feature in September, but, according to tests run by Tom's Hardware, it still might not be good enough.

The new version of Recall includes a sensitive information filter that's supposed to detect when there's information like credit card numbers and Social Security numbers on the screen. If it detects them, it will avoid taking a screenshot. When Tom's Hardware put this filter to the test, however, it failed in a number of situations.