Add another name to the list of games machines have dominated: Where’s Waldo.
Few things fill us with angst growing up quite as much as trying to find that candy cane man amid a scene of chaos. Most games get easier with practice, but lanky ol’ Wally just got harder to find with each turned page.
So you may watch a new robot called There’s Waldo with a mixture of respect and resentment as it locates our beanie-at-the-beach main character in under 4.5 seconds. It’s impressive, yes. But also a bit of a spoiler.
There’s Waldo is the brain child of Matt Reed, a creative technologist at the creative agency Redpepper. Reed and his colleagues built the bot out of a uArm Swift Pro that’s controlled by a Raspberry Pi computer. On the front of the arm sits a Vision Camera Kit, which snaps a photo of the puzzle and runs it through the computer vision program OpenCV to recognize faces in the crowd.
The secret sauce is in Google’s AutoML Vision, a user-friendly system that lets users train A.I. without prior knowledge about coding. Reed trained his algorithm on 62 distinct photos of Waldo he found online. That might not seem like a lot of data, given that A.I. are often trained on many times that, but it seems to work.
In a video posted online, There’s Waldo is shown solving Where’s Waldo scenes in just a few seconds. When tasked with finding Waldo, the robot snaps a photo, picks out the faces, and decides which ones are mostly like Waldo based on anything higher than 95 percent confidence. It then reaches out with a tiny silicone hand and selects the character for good measure.
There’s Waldo records times between 4 seconds and 30 seconds depending on the number of characters on the page.
Reed was apparently inspired to develop the system after seeing the Amazon Rekognition for celebrities.
“That got me wondering if detecting illustrated celebrities was even possible,” he told Digital Trends. “For some reason that thought combined with my childhood memories of looking for Waldo hours on end popped together and spawned this project.”
The whole thing took Reed about a week to program, which is about how long many of us usually spend trying to find Waldo in the first place.
Google has been finding more ways to make Gemini a part of your daily routine, from putting together playlists to ordering groceries through your favorite apps. Now it's expanding those capabilities to more services, with Google Search joining in, too.
The company has announced a new Connected Apps Platform that makes it easier for developers to bring their services to both Gemini and Search. Alongside this, Google is adding new partners to its growing lineup, bringing more of the apps people already use into its AI ecosystem.
In the broad context of AI and its perceived harms, health and medical risks often pop up in the conversation. The warnings even come from frontier AI labs, such as ChatGPT-maker OpenAI and Anthropic, both of which have warned about misusing AI to cook up bioweapons. Security experts have already spotted a few such incidents in the wild, especially with the usage of AI models without any solid guardrails.
Well, it appears that the risks run deeper. As per an analysis appearing in the British Medical Journal (BMJ), AI bots are now being deployed to influence public policies governing health risks to citizens. In this instance, an attempt was made using AI bots to thwart regulations that aim to restrict the sale of disposable vapes, all while disguising the manipulation efforts as a response from real humans.
What if checking your drinking water for dangerous chemicals were as simple as plugging a tiny gadget into your smartphone? Researchers in Japan have developed a compact sensor that could make that possible, potentially bringing sophisticated water contamination testing out of expensive laboratories and into people's hands.
Scientists at the Okinawa Institute of Science and Technology (OIST) have created a portable device capable of detecting extremely low concentrations of PFAS, commonly known as "forever chemicals." The findings were published in the journal Biosensors and Bioelectronics, which demonstrate how an inexpensive sensor and a smartphone could make identifying potentially harmful chemicals in water a lot easier.