Redmond launches bot to caption photos online

Apr 18, 2016 12:11 GMT  ·  By

The Internet broke down Microsoft’ Tay chatbot, so the company wanted to give another chance to artificial intelligence with a similar technology, this time, support to caption photos submitted by users.

Called CaptionBot, Microsoft’s new service is supposed to determine the objects or persons showing up in photos, along with emotions thanks to Microsoft’s Computer Vision API, Emotion API, and Bing Image API.

“I was created to showcase some of the new capabilities of Microsoft Cognitive Services. These new capabilities are the result of years of research advancements. Specifically, I use Computer Vision and Natural Language to describe contents of images. I am still learning, so sometimes I get things wrong,” Microsoft says in the description of the bot.

Still failing to work properly

Unsurprisingly, CaptionBot still fails to be 100 percent accurate, and while it sometimes manages to detect the persons or objects in a photo, it also happens for the new AI project to fail in the most humorous ways.

Twitter users were the first to spot these failures, and @davidsim has tweeted a photo of Barack Obama dancing with Michelle Obama that CaptionBot mistakenly detected as a “a man in a suit and tie talking on a cell phone.” So, yes, CaptionBot says Michelle Obama is nothing more than a cell phone.

We’ve tested CaptionBot with a photo of Zlatan Ibrahimovic, and the result was pretty surprising. “I am not really confident, but I think it's a man holding a frisbee and he seems 😠.”

The good thing is that Microsoft wanted to make sure that CaptionBot doesn’t become a Hitler fan as his sister Tay, so the company implemented some limitations to block it from captioning such photos. It doesn’t work with Osama Bin Laden photos either, and most likely, there are more limitations to make sure that the bot learns how to behave.