Luckily, there is a subreddit for that. The What Is This Thing community is fortunately still going strong with more than 2.7 million members! It’s a place where you can post a picture of an item that you can’t for the life of you figure out what it is or where it came from. The Internet will most likely come to your aid, and the subreddit’s detectives will tell you what your strange object is. This time, we’ve compiled a new selection of the best entries from the community for you to enjoy. So scroll away and let us know your favorites by upvoting them! Bored Panda reached out to Celia Quillian, an AI expert who previously lent her expertise to TIME magazine and the Today Show. We wanted to know whether the current AI technology could make the subreddit’s job easier. She told us more about how AI models recognize objects in pictures, what challenges we might run into, and what the future of AI image recognition technology is. Read our interview with her below! More info: What Is This Thing? | Celia Quillian on Instagram | Celia Quillian on TikTok Redditors love to share their knowledge, and it’s especially rewarding when knowing something niche helps someone else. “Many [people] have unique and obscure knowledge and they like to use it. Get enough of them together and they cover most things,” the moderator told us a few years back. It’s hard not to get involved when some interesting items show up – that’s another reason for the subreddit’s popularity. On a more serious note, there’s also an affiliated subreddit called Trace an Object. Internet sleuths can put their skills to good use by searching for items posted by Europol and Australia’s “Trace an Object” websites or the FBI’s “Endangered Child Alert Program.” “We’ve had some luck in identifying some things, but there’s more yet unsolved and more to come as Europol does release some new information every few months,” the community’s moderator told us in 2020. Celia tells us that current AI models are capable of helping us identify an unknown object in a picture. “While at this point in time there are no AI models that can identify absolutely everything, there are many AI tools that have image recognition features,” Quillian explains. The best-known tools for that include ChatGPTPlus, Google Gemini, and Microsoft Copilot. “I’ve used these myself to help identify photos of plants or animals on nature walks,” Celia adds. “Say you made a unique object at home out of a variety of things around the house and posted it to the Internet saying ‘This invention is called a Riesmatch!’ If you were to then upload a photo of the ‘Reismatch’ into an AI tool and ask it to identify it, it would not be able to come up with the ‘correct’ name but rather suggest a description of all of the recognized objects it is made of.” And some companies are already doing that! Like the Lithuanian startup Oxipit, which creates AI tools for global medical imaging. Companies that use this technology might help in addressing the global radiologist shortage and improve diagnostic quality. “I think one of the larger risks, which can be mitigated, is the inherent bias the training data set could have for all use cases, without the proper bounds in place,” Quillian adds. Follow Bored Panda on Google News! Follow us on Flipboard.com/@boredpanda! Please use high-res photos without watermarks Ooops! Your image is too large, maximum file size is 8 MB.
