Earth's'Gateway to Hell' keeps getting bigger Breakthroughs, discoveries, and DIY tips sent six days a week. In a remote area of the Siberian tundra, there's a place that locals call "The Gateway to Hell." In the summer, its peaceful waterfall sounds are interrupted by the booms and crashes of falling earth. The indigenous Yakut people are rightfully wary of the massive, slowly collapsing crater. Crater is an unsettling mark of our changing world.

'Thank God they're still alive': Kaiser therapists claim its new screening system puts patients at higher risk by delaying their care Kaiser pushed back on striking workers' claims and AI fears, saying it delivers'timely, high-quality care to meet members' needs' I lana Marcucci-Morris is worried about the patients she treats and how long it took for them to arrive in her office. At Kaiser Permanente's psychiatry outpatient clinic in Oakland, California, she says she increasingly finds herself assessing people experiencing severe mental health issues whom she believes should have been sent to the emergency room weeks earlier. For those who do make it to their appointments, she thinks: "Thank God they're still alive." It wasn't always this way, according to Marcucci-Morris, a licensed clinical social worker. Licensed professionals used to almost always be the first point of contact for patients with behavioral health issues at Kaiser, she said. She has noticed a change since January 2024, after the healthcare giant introduced a new screening process for first-time patients.

How the data center boom plunged one small Pennsylvania town into chaos. Valley View Estates is set to be surrounded by data centers. Get your news from a source that's not owned and controlled by oligarchs. "I don't like to see anyone upset," said Nick Farris of Provident Real Estate Advisors. He was sitting in the front of a crowd of roughly 150 inside Valley View High School's auditorium in Archbald, a town of about 7,500, huddled between two mountain ranges in Pennsylvania's Lackawanna Valley. Farris was there to represent the developer for Project Scott, one of many data center campuses coming to town. "I think that this is the best data center site in this area of the country, by far." The audience had been fairly quiet, bundled in thick coats against the late January cold. But as Farris spoke about data centers as a boon for communities, they began to laugh, drawing a rebuke from town officials. "What about the children?" someone shouted from the crowd. The children were watching from the walls; long banners of Valley View Performing Arts students hanging around the auditorium like championship pennants. Project Scott and four other data facilities will sit just a few thousand feet from the middle and high schools. He was referring to Lockheed Martin's 350,000-square-foot Missiles and Fire Control facility directly next to the high school, parts of which are highly contaminated . "That sucks too!" another attendee yelled back.

Trump administration'unlocks' 140MILLION barrels of precious Iranian oil with major policy change to fight back against'hoarding' China... here's what it means for your wallet Buffy the Vampire Slayer star Nicholas Brendon dead at 54 as'heartbroken' family reveal cause of death Joseph Duggar's wife Kendra is arrested for allegedly endangering welfare of a minor as he faces new charges Behind closed doors, the Duggar family's next nightmare began long before Joseph's arrest: Insiders reveal what they knew and how they plan to recover America is about to be torn apart by a financial tsunami - and it's not just an oil crisis to fear. However, it seems not every corner of Britain is quite so beautiful - as a survey has revealed the least scenic locations. Voters on the Scenic Or Not survey awarded the top spot to Basingstoke's Newbury Road. This unappealing location received the lowest possible score, with just one out of 10 for'scenicness'. And while Cornwall might be renowned for its beautiful scenery, a rather less attractive part of the county - the Electricity Station in Landulph - joins Basingstoke at the bottom of the pile.

Will Gulf states join war? One police officer has been killed in a drone strike by "outlaw groups" on the headquarters of the Iraqi National Intelligence Service in the heart of capital Baghdad. "A drone targeted the headquarters of the Iraqi National Intelligence Service in the Mansour district" at about 10am local time (07:00 GMT), General Saad Maan, head of the Iraqi government's security media unit, said in a brief statement on Saturday. Another drone, filming the operation, crashed into a private members ' sports club popular with the Iraqi elite and foreign diplomats, according to the same source. The drone attack on the headquarters of the National Intelligence Service came hours after another attack on the US military complex.

Two people were killed in a Russian drone attack on a home in the Ukrainian city of Zaporizhzhia, local authorities say. Two children, 11 and 15, were also injured in the attack which took place on the eve of new talks between Ukrainian and American negotiators in the US. Negotiations on ending the war have been on hold since the start of the latest conflict in Iran. President Volodymy Zelensky wants his negotiators to discuss the US decision to ease sanctions on Russian oil - implemented by Washington to help keep down global energy prices. Talks mediated by the US have so far failed to stop the fighting in Ukraine or change Russia's demands, and there is little hope of a breakthrough.

The Amazon Prime prank series amplifies the hijinks of workplace dynamics, while showing how people find purpose--and community--in their jobs despite impossible situations. Anthony Norman is your typical Gen Z worker: 25, a little wayward, and struggling to find a full time job. Unemployment rates are high . AI is creating a crisis for young people trying to enter the workforce. And several companies--including Amazon, Block, and Meta --have embraced tech's latest era of layoffmaxxing, with some cutting their staff by 20 percent.

As large language models (LLMs) become increasingly prevalent across many real-world applications, understanding and enhancing their robustness to adversarial attacks is of paramount importance. Existing methods for identifying adversarial prompts tend to focus on specific domains, lack diversity, or require extensive human annotations. To address these limitations, we present Rainbow Teaming, a novel black-box approach for producing a diverse collection of adversarial prompts. Rainbow Teaming casts adversarial prompt generation as a quality-diversity problem and uses open-ended search to generate prompts that are both effective and diverse. Focusing on the safety domain, we use Rainbow Teaming to target various state-of-the-art LLMs, including the Llama 2 and Llama 3 models. Our approach reveals hundreds of effective adversarial prompts, with an attack success rate exceeding 90% across all tested models. Furthermore, we demonstrate that prompts generated by Rainbow Teaming are highly transferable and that fine-tuning models with synthetic data generated by our method significantly enhances their safety without sacrificing general performance or helpfulness. We additionally explore the versatility of Rainbow Teaming by applying it to question answering and cybersecurity, showcasing its potential to drive robust open-ended self-improvement in a wide range of applications.

Vision-Language Models (VLMs) excel in generating textual responses from visual inputs, but their versatility raises security concerns. This study takes the first step in exposing VLMs' susceptibility to data poisoning attacks that can manipulate responses to innocuous, everyday prompts. We introduce Shadowcast, a stealthy data poisoning attack where poison samples are visually indistinguishable from benign images with matching texts. Shadowcast demonstrates effectiveness in two attack types. The first is a traditional Label Attack, tricking VLMs into misidentifying class labels, such as confusing Donald Trump for Joe Biden.

The Department of Defense alleges the AI developer could manipulate models in the middle of war. Company executives argue that's impossible. Anthropic cannot manipulate its generative AI model Claude once the US military has it running, an executive wrote in a court filing on Friday. The statement was made in response to accusations from the Trump administration about the company potentially tampering with its AI tools during war . "Anthropic has never had the ability to cause Claude to stop working, alter its functionality, shut off access, or otherwise influence or imperil military operations," Thiyagu Ramasamy, Anthropic's head of public sector, wrote .