Well File:

All the AI news of the week: ChatGPT debuts o3 and o4-mini, Gemini talks to dolphins

Mashable

Just like AI models, AI news never sleeps. Every week, we're inundated with new models, products, industry rumors, legal and ethical crises, and viral trends. If that weren't enough, the rival AI hype/doom chatter online makes it hard to keep track of what's really important. But we've sifted through it all to round up the most notable AI news of the week from the heavyweights like OpenAI and Google, as well as the AI ecosystem at large. As of this writing, the popular AI leaderboard LMArena ranks Gemini 2.5 Pro as the model to beat, followed by ChatGPT 4o, and Grok-3 Preview.


Google rolls out Gemini Live screen sharing to all Android users

Mashable

Google's Gemini Live launched for all Android users over the weekend. That means far more folks now have access to the AI assistant's tool that lets users live-share a video or screen. The feature rolled out last month to Pixel 9, Galaxy S25, and Gemini Advanced subscribers before now going wide to all Android users, Android Central reported. The move was expected: Google noted earlier this month that the feature was rolling out to all Android users with the Gemini app. The idea behind the tool is that the AI assistant sees what you see, either via your camera or on your screen.


New PR? Humanoid robots in China competed in their first half-marathon

Mashable

Over the weekend, humans running as fast as they could were chased by robots through the streets of Beijing, China. To be more specific, it was a half-marathon race, and the robots lagged far behind the humans. On Saturday, China held what it's calling the world's first humanoid half-marathon. Over 20 two-legged humanoid robots competed alongside real human runners, according to state-run news outlet Beijing Daily, via CNN World. The teams were from Chinese universities and companies publicizing their humanoid robotics advancements, which China's Ministry of Industry and Information Technology has dubbed a critical area for competing with the U.S. As CNN reports, local governments in major cities like Beijing, Shanghai, and Shenzhen have invested an estimated 10 billion in developing humanoid robotics to compete with humanoids from U.S. rivals like Boston Dynamics, Figure AI, and Elon Musk's Tesla.


Wikipedia has a solution for the deluge of AI training bots hogging its servers

Mashable

You're not the only one who turns to Wikipedia for quick facts. Lately, a deluge of AI bots training on Wikipedia articles has put enormous strain on the organization's servers. To curb the influx of "non-human traffic" scraping the site for training data, Wikipedia is taking a proactive approach: serving up its data directly to AI developers. On Wednesday, the Wikimedia Foundation announced a partnership with Google-owned company Kaggle to release a beta dataset "featuring structured Wikipedia content in English and French." Uploaded on April 15, the company said the dataset "simplifies access to clean, pre-parsed article data that's immediately usable for modeling, benchmarking, alignment, fine-tuning, and exploratory analysis."


OpenAIs o3 and o4-mini hallucinate way higher than previous models

Mashable

By OpenAI's own testing, its newest reasoning models, o3 and o4-mini, hallucinate significantly higher than o1. First reported by TechCrunch, OpenAI's system card detailed the PersonQA evaluation results, designed to test for hallucinations. From the results of this evaluation, o3's hallucination rate is 33 percent, and o4-mini's hallucination rate is 48 percent -- almost half of the time. By comparison, o1's hallucination rate is 16 percent, meaning o3 hallucinated about twice as often. The system card noted how o3 "tends to make more claims overall, leading to more accurate claims as well as more inaccurate/hallucinated claims." But OpenAI doesn't know the underlying cause, simply saying, "More research is needed to understand the cause of this result."


We tried the ChatGPT reverse location search trend, and its scary

Mashable

ChatGPT users have discovered that the popular AI chatbot can serve as a reverse-location search tool. In other words, you can show ChatGPT a picture, and it can pretty reliably tell you where it was taken. The trend is inspired by the online game Geoguessr, where folks try to figure out a location from a simple web image. We decided to put this new ChatGPT trend to the test, and the results were downright scary. Mashable tech reporters prompted ChatGPT to play a geo-guessing game and uploaded a series of photos.


How Instagram is using AI to uncover teen accounts lying about their age

ZDNet

If your teen has a fake birthday on their Instagram profile to get around age restriction policies, their days of beating the system may soon be over. In a blog post, Instagram says that starting today it will use AI to detect users lying about their age and automatically move those accounts to one of the limited teen accounts that debuted last fall. First, it will monitor which profiles and content an account interacts with. Since people in the same age range generally enjoy similar content and interact with each other, if a non‑teen account interacts heavily with teen accounts and teen‑related content, Meta may flag that account. Second, Meta will review what it calls "strong signals of age," or things like birthday messages -- for example, if another user posts something like "Screaming happy 15th birthday to my best friend."


Take AI-powered piano lessons for life for 50% off

Mashable

TL;DR: Skoove Premium Piano Lessons uses advanced AI to give you curated virtual piano lessons, and right now a lifetime subscription can be yours for just 113.07 (reg. Whether you've dabbled in lessons as a kid or never sat on a piano bench, Skoove Premium Piano Lessons can help you master the keys from the comfort of home. All you'll need is a tablet, a keyboard, and this AI-powered app. Right now, you can save 50% on a lifetime subscription and keep honing your craft for life for just 113.07 (reg. Skoove offers AI-powered piano lessons that let you tickle the ivories in your spare time.


Aiper Scuba X1 review: If looks could clean your pool

PCWorld

The Aiper Scuba X1 looks--and is priced--like a high-end robotic pool cleaner, but it's a weak performer and it's a bear to clean after a session in the pool. Aiper makes some excellent robotic pool cleanrs--such as its stellar workhorse, the Seagull Pro--but it also has a few duds in its arsenal, including the Seagull Plus and the Scuba S1. With its latest robot, the Scuba X1, Aiper looks to bring some higher-end features (including smart connectivity) to the lineup. With a street price of 1,200, it's one of Aiper's most expensive models–and it's got the gold trim to prove it. The Aiper Scuba X1 doesn't change the basic design that most of Aiper's full-size robots have followed for years: Compare its design to the aforementioned Seagull Pro, Seagull Plus, and Scuba S1.


When Google and ChatGPT translated my image text, one failed - and the other got weird

ZDNet

One of the fun aspects of my roles as technical columnist and YouTube producer is testing new gadgets as they come out. I've been testing an Anycubic Kobra 3 3D printer for a while now, which led to this article. An unfortunate new trend is that most of the bigger 3D printer companies have taken an open-source slicer called Orca Slicer and rebranded it for their own use, adding machine-specific code to enable printing with their printers. Also: This ChatGPT trick can reveal where your photo was taken - and it's unsettling Anycubic did this with their Anycubic Slicer Next, which substantially improved their previous home-grown slicer. Anycubic is based in China.