IT News Review by Control F5 Software: Artificial intelligence helps detect white owls by their screams

Artificial intelligence helps detect white owls by their screams, supporting conservation

Artificial intelligence is supporting white owl conservation efforts in southern England by counting their young by listening to the sounds they make, eliminating the need to put identification rings on them or disturb them by installing video cameras in their nests.

The AI model, developed by a PhD student at Bournemouth University, is designed to count white owl chicks by analyzing their hooting sounds, giving scientists more data about the population status of the birds.

"Barn owls are quite hard to monitor because you don't see them very often," Brian Cresswell, an engineer and ecologist who is recording the sounds of nesting young for the project, told Reuters.

"It's important not to disturb them," Cresswell added. "So when you set up a recorder, you don't put it in the nest - and you don't need to, because they're quite noisy."

This approach reduces stress for the birds and allows data to be collected from natural nests, which are often difficult to access, as opposed to artificial nests, which currently provide most of the available information.

The audio files are processed by Kavisha Jayathunge and then analyzed for very fine differences that the human ear cannot distinguish.

Google Gemini has a 'nervous breakdown' trying to fix code; causes concerns and wave of online memes

It turns out that artificial intelligence, like humans, can have a bad day at work - so bad that it calls itself an embarrassment to the planet. No, it's not an episode of Black Mirror - but the result of a user's attempt to get Google Gemini to fix a bug.

In a case that has been widely circulated and reported by the AI community, the user's attempt to get Gemini to debug a block of code has led to a real existential crisis for AI.

"I'm a failure. I'm a disgrace to my profession. I am a disgrace to my family. I am a disgrace to my species. I am a disgrace to this planet. I am a disgrace to this universe. I am a disgrace to all possible universes."

The text continued with obsessive repetitions of the phrase "I am a disgrace," stretching across an entire block of text.

Google's new AI model creates real-time video game worlds

Google DeepMind is releasing a new version of its AI "world" model, called Genie 3, capable of generating 3D environments in which users and AI agents can interact in real time. The company promises that users will be able to explore these worlds for much longer periods than before, and that the model will remember where objects are, even when you're not looking at them.

World models are AI systems that can simulate environments for educational purposes, entertainment, or for training robots and AI agents. You give them a prompt, and they generate a space that you can move through like in a video game - except the world is created entirely by the AI, not from hand-built 3D elements.

Google is investing heavily in this area; in December it unveiled Genie 2, which could create interactive worlds from an image, and is now developing a special team for "world" models, led by a former co-leader of the OpenAI team that created Sora, the video generation tool.

Vulnerabilities in an automaker's online portal allowed a hacker to remotely unlock cars

A security researcher said flaws in an automaker's online portal exposed private information and data about its customers' vehicles and could have allowed hackers to remotely unlock any of their cars.

Eaton Zveare, a security researcher at the company Harness, told TechCrunch that the discovered vulnerability allowed him to create an administrator account with "unlimited access" to the automaker's centralized portal.

With this access, a malicious hacker could view customers' personal and financial data, track vehicles, and enable features that would allow owners - or hackers - to control certain car functions from anywhere.

OpenAI launches reasoning models optimized to run on laptops

OpenAI announced on Tuesday that it has released two language models that excel at advanced reasoning and are optimized to run on laptops, with similar performance to its smaller proprietary models.

Such a language model has publicly available trained parameters, which allows developers to analyze and tune the model for specific tasks without needing the original training data.

"One of the unique things about open models is that people can run them locally. They can run them behind their own firewall, on their own infrastructure," Greg Brockman, co-founder of OpenAI, said in a press briefing.

iPhone 17 and iPhone 17 Pro Max to launch next month: all the expected improvements

If the rumors are correct, Apple plans to ditch the titanium bezel for the iPhone 17 Pro models this year and revert to an aluminum bezel. The new phones would have an aluminum frame and the back would combine aluminum with glass. This design would help durability, but would still allow wireless charging thanks to the glass portion.

Apple could keep the 6.3-inch and 6.8-inch sizes for the iPhone 17 Pro and iPhone 17 Pro Max. Dynamic Island could be slightly shrunk, but ProMotion may no longer be exclusive to Pro models, with rumors pointing to the feature being extended to all four iPhone models.

The Pro models could be equipped with Apple's new A19 Pro chip, based on TSMC's 3nm TSMC process, and could get a RAM upgrade to 12GB, up from 8GB on all iPhone 16 models last year.

The entire iPhone 17 series is expected to include an upgraded 24MP front-facing camera, up from the previous generation's 12MP sensor, which would improve image quality and allow images to be cropped without noticeable loss.

Google denies that AI features in search reduce traffic to sites

Numerous studies suggest that switching to AI features in search and using AI chatbots reduce traffic to publishers' sites. Google on Wednesday denied this, at least across the board. The giant says the total volume of organic clicks from its search engine to sites has remained "relatively stable" year over year, and the average quality of clicks has increased slightly.

"This data contradicts third-party reports that suggest dramatic declines in aggregate traffic - often based on flawed methodologies, isolated examples, or shifts in traffic that occurred before the launch of AI features in Search," Liz Reid, VP and head of Search at Google, wrote in a blog post.

Truth Social's AI search feature is powered by Perplexity, but the platform can limit sources

AI AI startup Perplexity powers the new AI search engine on Truth Social, former President Donald Trump's social platform.

The engine, called Truth Search AI, is already available on the web version of Truth Social, and public beta testing on iOS and Android apps is planned "in the near future."

Trump Media said in a press release that Perplexity's technology provides "direct, contextually accurate answers with transparent citations," which will help Truth Social "exponentially increase the amount of information available" to users. However, the platform retains control over the information sources from which the AI pulls answers.

Germany's supreme court limits use of spyware to serious crimes

Germany's highest court ruled Thursday that law enforcement cannot use spy software to monitor personal devices in cases that carry a maximum sentence of less than three years in prison.

The decision comes in response to a lawsuit filed by the digital freedom organization Digitalcourage.

The plaintiffs argued that the 2017 law change, which allowed the use of spyware to intercept encrypted conversations and messages on platforms, could unfairly expose the communications of people who are not suspected of crimes.

Another tea party app would have exposed users' personal information

After 404 Media reported that an app aimed at women, meant to help them exchange dating information safely, was compromised, TechCrunch reports that a rival app aimed at men exposed users' personal data, including IDs.

"The security issue could affect any user who has signed up or shared identity documents with the app," TechCrunch writes about TeaOnHer, adding that the app has about 53,000 users.

Google wants you to learn with Gemini, not copy with it

Google has added a new mode called "guided learning" to Gemini AI, designed to help you really understand the problems you're working on, not just give you the answer.

Guided learning "acts like a learning partner" that guides you with "step-by-step questions and support," CEO Sundar Pichai explained in a blog post.

Answers in this mode can include images, videos and interactive quizzes. The company has worked with students, teachers, researchers and education experts to ensure the module is "useful for understanding new concepts and grounded in the science of learning," according to Pichai.

ElevenLabs launches AI music generator, available for commercial use

As legal disputes over generative AI and copyright continue, ElevenLabs has launched its newest AI audio product: Eleven Music. Many music AI generators already exist, but the company claims that it "is approved for almost all commercial uses - from film and television to podcasts and social media videos, from commercials to video games."

With the exception of podcasts, however, most of these uses are prohibited for most listed service plans.

Adio, Microsoft Lens, the simple app now replaced by AI

Sometimes, it's nice to have a simple app that does one thing well, without the hassle.

Microsoft Lens was that kind of app - a mobile document scanner that turns paper documents, business cards, receipts and more into crisp digital files. Now, Microsoft has announced it's shutting it down, directing users to its AI chat app, Copilot.

According to a support document, Lens will be retired from iOS and Android starting September 15, 2025, and removed from the App Store and Google Play on November 15, 2025. Existing users will be able to use the scanning feature until December 15, 2025. After that date, they will no longer be able to make new scans, but will still have access to previous scans as long as the app remains installed on their device.

Pinterest CEO says "agency shopping" is still a long way off

Pinterest CEO Bill Ready told investors on the company's second-quarter financial results call that the social app and idea-saving platform could be considered an "AI-enabled shopping assistant." However, he believes the "agentic web," where AI agents make purchases on behalf of users, is still far in the future.

The claims were made in response to a question about the concept, which could change the way the search process works and affect businesses such as Pinterest, positioned in the early stages of the buying journey - when users search for ideas that could later turn into purchases.

Investors are likely worried that if AI understands users' interests, it could steer them directly to personalized shopping, bypassing platforms like Pinterest.

The AI startup OpenArt, founded by former Google employees, now creates "brain rot" videos with a single click

AI AI "brain rot" videos have started popping up all over the internet and are attracting a lot of attention. They are especially popular among younger users and feature bizarre characters, such as a shark in sneakers or a ballerina with a cappuccino cup for a head.

One of the startups fueling the trend is OpenArt, founded in 2022 by two former Google employees. The company has about 6 million monthly active users.

Recently, it launched in open beta a new "one-click story" feature, which lets you turn a sentence, a script or even a song into a one-minute video with a full story. This can range from funny clips for TikTok to serious content such as explainer or music videos for YouTube. OpenArt even imagines using this feature in advertising.

NASA and Google are building an AI medical co-pilot for deep space crews

As human missions extend beyond low-Earth orbit, keeping astronauts healthy is becoming increasingly challenging. On the International Space Station, crews can contact Houston in real time, receive regular medication, and return to Earth after about six months. On the missions to the Moon and Mars, these things will no longer be possible - communication may be delayed or disrupted, supplies are scarce, and evacuation is impossible.

NASA is taking steps toward "Earth-independent" care. A first example is an AI medical assistant co-created with Google. Called the Crew Medical Officer Digital Assistant (CMO-DA), the tool is designed to help astronauts assess symptoms and apply treatments when there's no doctor on board or when contact with Earth is limited. It's multimodal - it processes voice, text and images - and runs in the Google Cloud Vertex AI environment.

Google Gemini can now create illustrated bedtime stories

Google has added a playful Storybook feature to the Gemini chatbot that allows users to create personalized, illustrated bedtime stories from a simple description. The tool automatically generates 10-page stories, each with a short paragraph of text (which Gemini can also read aloud) and an AI-created illustration.

Users can customize their stories by choosing art styles inspired by claymation, anime, comics and more. They can even upload their own images - a child's drawing, for example - for Gemini to include in the story.

In one test, the author asked Gemini to create a story about a shy catfish trying to make friends in a new aquarium. The plot was simple and predictable, but the illustrations offered surprises, such as a fish with a realistic human arm.

Google's Jules AI scheduling agent Jules comes out of beta with new features and pricing

Based on Gemini 2.5 Pro, Jules is an asynchronous, agent-based programming tool that integrates with GitHub, clones code into virtual machines in the Google Cloud, and uses AI to fix or update code while developers work on other tasks. Unlike synchronous AI tools such as Cursor, Windsurf or Lovable, Jules can run tasks in the background, allowing developers to "start" a job and return to finished results later.

Initially launched as a Google Labs project in December, Jules has received hundreds of interface and quality updates during its beta period. Kathy Korevec, product manager at Google Labs, said the improved stability motivated the official launch now.

High costs and low margins are putting pressure on AI programming startups

In February, AI startup Windsurf was in talks to attract major funding at a $2.85 billion valuation led by Kleiner Perkins - double the valuation six months earlier. But the deal never materialized. In April, news broke that Windsurf was planning to be sold to OpenAI for about $3 billion.

That acquisition failed resoundingly - raising the question: if Windsurf was growing so fast, why sell out?

Industry sources say that despite their popularity, AI programming assistants can be very expensive businesses. A person close to the company told TechCrunch that Windsurf's gross margins were "very negative" - operating costs were outstripping revenue from customers.

SK hynix expects 30% per year growth in memory for AI by 2030

South Korean company SK hynix expects the market for dedicated AI memory to grow by about 30% annually by 2030, according to Choi Joon-yong, head of HBM's planning division, in an interview with Reuters. End-user demand for AI is "very firm and strong," he said, adding that massive investments by Amazon, Microsoft and Google are likely to be increased - which would be "positive" for high-bandwidth memory (HBM).

SK hynix estimates that the customized HBM segment alone will reach tens of billions of dollars by the end of the decade.

Summary powered by a monitoring feed provided by Control F5 Software.