Published June 4, 2025 10:00am EDT close Windows bug leaves computer Wi-Fi vulnerable to hackers Kurt "CyberGuy" Knutsson explains how to keep your Windows computer safe and the security risks of online retail giant Temu. NEWYou can now listen to Fox News articles! Hackers are no longer targeting only tech giants or hospitals. Any business that collects valuable personal information, such as names, phone numbers, email addresses or even basic financial details, is now a target.Companies that rely heavily on third-party vendors or outsourced customer support are even more at risk, especially if they are not particularly strong in the technology sector.German retailer Adidas learned this the hard way. The company recently confirmed a data breach involving one of its external partners, and although it has acknowledged the issue, many important details are still missing. A hacker at work (Kurt "CyberGuy" Knutsson)Adidas confirms vendor breach: Here’s what we knowAdidas has officially acknowledged that a third-party vendor suffered a breach, resulting in unauthorized access to consumer data. In a public notice titled "Data Security Information," the company revealed that a "third-party customer service provider" had been compromised. While the brand was initially silent on the scope, it had already been reported earlier this month that customers in Turkey and Korea had received breach notifications.MASSIVE DATA BREACH EXPOSES 184 MILLION PASSWORDS AND LOGINSAdidas posted this information on both its German and English websites. However, no specific region or number of affected individuals has been confirmed. The company’s statement did clarify that no payment information, such as credit card details, nor passwords were included in the breach. Instead, it involved contact details submitted by users to Adidas’ help desk in the past.Data obtained reportedly includes names, phone numbers, email addresses and dates of birth. While this might seem limited compared to financial data, this type of information can be exploited for phishing scams and identity theft.  An Adidas sign (Kurt "CyberGuy" Knutsson)What Adidas told customers after the breachIn the wake of the breach, Adidas began notifying potentially affected customers directly. The company's email to customers below aimed to reassure recipients and clarify what information was involved. Here is the full text of the notification sent to affected individuals.Dear customer,We are writing to inform you of an issue that we recently became aware of which may have impacted some of your data.What happenedadidas recently learned that an unauthorized external party gained access to certain customer data through a third-party customer service provider.What information was involvedThe affected data does not contain passwords, credit card or any other payment-related information. Nor have any Social Security numbers been impacted.It mainly consists of contact information relating to customers who had contacted our customer service help desk in the past. This may have included one or more of the following: name, email address, telephone number, gender and/or birth date.What we are doing Privacy and the security of your data is our priority. Upon becoming aware of this incident, adidas took proactive and immediate steps to investigate and contain the incident. This includes further enhancing security measures and resetting passwords for customer service accounts.What you can doWe are currently unaware of any harm (such as identity theft or fraud) being caused to our customers as a result of this incident. There are no immediate steps that you need to take. Although, as always, please remain vigilant and look out for any suspicious messages. As a reminder, adidas will never directly contact you to ask that you provide us with financial information, such as your credit card details, bank account information or passwords.Who you can contactIf you have any questions, then please contact our Customer Service team at https://www.adidas.com/us/helpWe apologise for any inconvenience caused by this incident.adidas TeamWhat Adidas hasn’t said about the vendor hackDespite the official acknowledgment, several questions remain unanswered. Adidas has yet to clarify whether this is a single breach affecting multiple regions or several separate incidents. The lack of transparency around the name of the third-party vendor and the absence of concrete numbers or locations for affected users has created frustration among observers and possibly among customers themselves.The earlier regional reports from Turkey and Korea might suggest that this incident was either global in scale or that similar third-party vendors were independently targeted. In either case, the company's current handling of the situation has left room for speculation. Adidas claims it is in the process of informing potentially affected customers, but it has not detailed the method or timeline for this outreach.We reached out to adidas for a comment, and a representative referred us to this statement on their website. In part, the company said, "We remain fully committed to protecting the privacy and security of our consumers, and sincerely regret any inconvenience or concern caused by this incident."GET FOX BUSINESS ON THE GO BY CLICKING HERE An Adidas shoe (Kurt "CyberGuy" Knutsson)6 critical steps to take after the Adidas data breachIf you think you were affected or just want to be cautious, here are some steps you can take right now to stay safe from the Adidas data breach:1. Scrub your data from the internet using a personal data removal service: The more exposed your personal information is online, the easier it is for scammers to use it against you. Following the Adidas breach, consider removing your information from public databases and people-search sites. Check out my top picks for data removal services here.Get a free scan to find out if your personal information is already out on the web.2. Watch out for phishing scams and use strong antivirus software: With access to your email and phone number, Adidas attackers can craft convincing phishing emails pretending to be from healthcare providers or banks. These emails might include malicious links designed to install malware or steal login information. To defend yourself, use a strong antivirus program. Get my picks for the best 2025 antivirus protection winners for your Windows, Mac, Android and iOS devices.3. Safeguard against identity theft and use identity theft protection: Hackers now have access to high-value information from the Adidas breach. This makes you a prime target for identity theft. You might want to consider investing in identity theft protection, which can also assist you in freezing your bank and credit card accounts to prevent further unauthorized use by criminals. Signing up for identity theft protection gives you 24/7 monitoring, alerts for unusual activity and support if your identity is stolen. See my tips and best picks on how to protect yourself from identity theft.4. Set up fraud alerts: Requesting fraud alerts notifies creditors that they need extra verification before issuing credit in your name. You can request fraud alerts through any one of the three major credit bureaus; they’ll notify the others. This adds another layer of protection without completely freezing access to credit.5. Change passwords and use a password manager: Update passwords on any accounts tied to compromised data. Use unique passwords that are hard to guess and let a password manager do the heavy lifting by generating secure ones for you. Reused passwords are an easy target after breaches. Consider password managers for convenience and security. Get more details about my best expert-reviewed password managers of 2025 here.6. Be wary of social engineering attacks: Hackers may use stolen details like names or birthdates from breaches in phone scams or fake customer service calls designed to trick you into revealing more sensitive info. Never share personal details over unsolicited calls or emails. Social engineering attacks rely on trust, and vigilance is key.Kurt’s key takeawayThe Adidas breach shows that even companies with decades of brand equity and a massive global footprint are not immune to lapses in data security. It underscores the need for companies to go beyond basic compliance and actively evaluate the cybersecurity standards of every partner in their ecosystem. Consumers are becoming increasingly aware of the trade-offs they make when sharing their personal information, and brands that fail to meet this moment may find their reputations eroding faster than they expect.CLICK HERE TO GET THE FOX NEWS APPShould retailers be penalized for neglecting basic cybersecurity practices? Let us know by writing us atCyberguy.com/Contact.For more of my tech tips and security alerts, subscribe to my free CyberGuy Report Newsletter by heading to Cyberguy.com/Newsletter.Ask Kurt a question or let us know what stories you'd like us to cover.Follow Kurt on his social channels:Answers to the most-asked CyberGuy questions:New from Kurt:Copyright 2025 CyberGuy.com. All rights reserved. Kurt "CyberGuy" Knutsson is an award-winning tech journalist who has a deep love of technology, gear and gadgets that make life better with his contributions for Fox News & FOX Business beginning mornings on "FOX & Friends." Got a tech question? Get Kurt’s free CyberGuy Newsletter, share your voice, a story idea or comment at CyberGuy.com.

Photo Credit: Yunnan Institute of Cultural Relics and Archaeology Ancient DNA from Yunnan Uncovers Ghost Lineage Linked to Tibetan Ancestry Highlights 7,100-year-old Xingyi_EN carries DNA from a mysterious ghost lineage Study links ancient Yunnan DNA to modern Tibetan ancestry Central Yunnan ancestry tied to early Austroasiatic populations Advertisement A new study on a 7,100-year-old skeleton from China has revealed a "ghost" lineage that only existed in theories until now. Skeleton of the early Neolithic woman, known as Xingyi_EN, unearthed at the Xingyi archaeological site in southwestern China's Yunnan province. Her DNA links her to a deeply divergent human population that may have contributed to the ancestry of modern Tibetans. This study also reveals a distinct Central Yunnan ancestry connected to early Austroasiatic-speaking groups. This discovery makes Yunnan as a key region to understand the ancient genetic history of East and Southeast Asia. The detailed analysis of 127 human genomes from southwestern China is published in a study in the journal Science.Xingyi_EN: A Genetic Link to a Mysterious PastAccording to the study, radiocarbon dating indicates Xingyi_EN lived around 7,100 years ago and isotope analysis suggests she lived as a hunter-gatherer. Genetic sequencing revealed her ancestry from a deeply diverged human lineage—now named the Basal Asian Xingyi lineage. This lineage diverged from other modern human groups over 40,000 years ago and remained isolated for thousands of years without mixing with other populations.This "ghost" lineage does not match DNA from Neanderthals or Denisovans but appears to have later contributed to the ancestry of some modern Tibetans. Xingyi_EN represents the first physical evidence of this previously unknown population.Yunnan's significance as a reservoir of deep human diversityMost of the skeletons that the researchers sampled were dated between 1,400 and 7,150 years ago and came from Yunnan province, which today has the highest ethnic and linguistic diversity in all of China."Ancient humans that lived in this region may be key to addressing several remaining questions on the prehistoric populations of East and Southeast Asia," the researchers wrote in the study. Those unanswered questions include the origins of people who live on the Tibetan Plateau, as previous studies have shown that Tibetans have northern East Asian ancestry. For the latest tech news and reviews, follow Gadgets 360 on X, Facebook, WhatsApp, Threads and Google News. For the latest videos on gadgets and tech, subscribe to our YouTube channel. If you want to know everything about top influencers, follow our in-house Who'sThat360 on Instagram and YouTube. Further reading: Ancient DNA, Yunnan, Ghost Lineage, Tibetan Ancestry, Human Evolution, Xingyi_EN, Paleogenetics, Archaeogenomics, Prehistoric Asia, Science Study Gadgets 360 Staff The resident bot. If you email me, a human will respond. More Related Stories

The big Leslie Benzies interview: MindsEye, Everywhere, and the double-edged sword of GTA How Build A Rocket Boy developed its debut project Feature by Samuel Roberts Editorial Director Published on May 30, 2025 As the producer behind the Grand Theft Auto games from GTA 3 through to GTA 5, as well as Red Dead Redemption and LA Noire, any project with Leslie Benzies' name on it is going to be a lightning rod for attention. MindsEye, the first game from Benzies' studio Build A Rocket Boy, is getting plenty of it – even if some of that attention has been less positive. MindsEye is a single-player third-person shooter with vehicle gameplay, set in a Las Vegas-style city called Redrock. It's a techno-thriller story about a former soldier called Jacob Diaz – but it's clear from visiting BARB in Edinburgh this week that the game is envisioned as a gateway into something much larger, both in the fiction of MindsEye, and for players who pick the game up. That includes a user-generated content platform called Build.MindsEye, where players on PC can create levels using relatively straightforward tools that incorporate any object in the game. When asked if third-person shooter levels or driving sections were the limits of the build side of MindsEye, the developers showed other examples of how they can be used, like massively increasing the proportions of a basketball, dropping it into the world, and functionally making an in-game version of Rocket League. Still, while MindsEye launches on June 10, 2025, for PC and consoles, many questions remain unanswered, including the future of its long-gestating Everywhere project. Benzies sat down with GamesIndustry.biz earlier this week to talk us through his vision for the game. This interview has been edited for brevity and clarity. Image credit: Austin Hargrave What's your grand vision for MindsEye? What will it be at launch, and where is it going in the future? MindsEye is one story in an epic universe. The other stories take place at different time periods, and [at] different locations in the universe. This story is Jacob Diaz's story. There are also other stories within MindsEye, so we tell the backstories of other characters Jacob will meet. That's the way we're going to fill out the universe over time – so when you travel around, all the stories will be connected by one overarching theme, and each story will have different mechanics. And we'll give these mechanics to players within the creator tools. What will happen with the game after launch? [The studio] will support the game through Play.MindsEye, with continuous new content. Some of the content, like races, are made just for fun. But [with] most of the content, we'll try and incorporate it into the story. So once you've played the big overarching ten-year plan, you'll have a very good idea of what this universe looks like. We have plans to add multiplayer, [and] we have plans to make a full open world. And of course, we've also got to look at what players are creating, and incorporate that into our plans. Given the ease of the tools, we think there's going to be a high percentage of players who will jump in and give it a pop, see how it feels. Hopefully some will create compelling content we can then promote and make that part of our plans to push to other players. Is it best to think of MindsEye as the first game in a series of games? Or one game as part of a larger experience? MindsEye sits bang in the middle of our story. So, we're going to go back 10,000 years, then we're going to go forward a certain amount of time. It's the relevant piece of the puzzle that will have players asking questions of what the bigger story is. We've intentionally not released footage of huge parts of the game, because we don't want to spoil anything for players. But this story does take some unusual twists. What's your vision for the multiplayer component of the game? I guess there's two sides to the answer. The dream from the building side is to allow players the opportunity to create their own multiplayer open world games with ease. So anyone could pick up the game, jump in, drive around, stop at a point where they see something of interest, build a little mission, jump back in the car, drive again, build another mission. Once you've built a couple of hundred of these, you've built your own open world game. So, that's the build side. From our side, we want to [create] a place where people can socialise, play together, and engage in the stories that we build. So, we do have plans next year to launch an open world multiplayer game that takes place a year after MindsEye finishes. In the interim, we also have an open world free roam game that spans from when MindsEye finishes to the launch of the open world multiplayer game. All of these stories interconnect in a fairly unique and original way, which I think players like these days. They like the complexity of deeper stories. You're selling the base game at launch, with a pass for upcoming content additions. Do you have a vision for how you're going to package future stories in the overarching MindsEye experience? It depends on the scale of the story. Some will be free, and some will be paid. After you left Rockstar Games, what came next? What led to you building the studio? I spent a few years looking into some other things: going [into] some property development. Using some of the games experience, we made a thing called VR-Chitect, which allowed you to build houses and view them in VR. I spent a lot of time in Los Angeles at this point, and this is when the droughts were very bad. I got into [making] these machines that would suck water out of the air. Still sitting in my back garden in Los Angeles is this big clunky machine, it works like an air-conditioning unit. It could suck up one thousand litres of water. So I got involved with that. But there's really nothing like making games. The different types of people – the lawyers, the accountants, the programmers, the artists, the dancers, the singers – that bunch of people in one big pot, all working together, and turning something from a piece of paper into [something on the] screen – that's where I get my excitement. Since I was a kid, that's what I've wanted to do. I thought, 'I better get back into making games' because nothing else was as much fun. What was the journey towards creating MindsEye as your first standalone release? Your first game's always your hardest. You have to build systems, you have to build the team. Everything is new. You don't really see a lot on the screen until way down the line, because you're building underlying systems, physics systems, the gameplay systems. It's a slow start, but what you end up with is an engine, and obviously we use Unreal, which provides a certain level of support and building. On top of that, we've got to build our own stuff. [Plus], we have to pack up everything we build and present it nicely for the creator tools. So it adds this extra layer of complexity to everything. But now, given where we are, the speed that we can iterate, we can very quickly place enemies, place vehicles, place puzzles, whatever, and get a feel for a game. We've now got a great, experienced team – a lot of talented guys in there. In the old days, you'd get a game, stick it on the shelf, and you'd wave goodbye. It's not like that anymore. You're continually fixing things. When you release a game, you've suddenly got, not a hundred testers, but hopefully millions of testers. You've got to continually fix, continually optimise, and especially with the tools that we've got, we want to continually create new content. So MindsEye is a standalone game, and Everywhere is not mentioned anywhere on the Steam page. But obviously there's a strong 'build' component to this game, which was part of the Everywhere pitch. What does this mean for Everywhere, and what was behind the decision to package the game this way? This is all part of a bigger story and ecosystem that we've got planned. Everywhere is going to show up again pretty soon. Everything we're working on, there's a story behind it – a big overarching story. So Everywhere will come back, and it fits into this story somewhere. I can't tell you [where], because it would be a spoiler. But that's going to reappear soon, and it will all be a part of the same product. "I'm not sure it would've been smart as a company to say, 'we are going to compete with the biggest game on the planet'" Leslie Benzies, Build A Rocket Boy In terms of the tools, the tool doesn't really care what world you're building in. It sits separately. So any game we create, it will naturally work on top of it. But we're big fans of keeping everything thematically connected, or connected through a narrative, and you'll see it. The bigger story will become obvious, once you've played through all of MindsEye. Then you might start to see how it all connects together, to the Everywhere world. Has the landscape for something like Everywhere, or the build component to MindsEye, changed as platforms like UEFN have taken off or Roblox has become so huge? It's great to see these tools being used by people. I build a lot with my son, and when he builds, I see the excitement he gets. It reminds me of when I was a kid with my Dragon 32 computer, managing to get a little character moving on the screen – that excitement of, 'wow, I did that'. Giving that to other people is massive. It's still very difficult to build in Roblox. For example, when my son wants to do it, I have to jump in. I used to be a programmer, and I struggle to build in there. When he wants to run around and scream with his friends he's in Roblox; when he wants to build he'll jump into Minecraft, because Minecraft is a much easier system to build within. And I think we sit somewhere in the middle: you can get very high quality, fun games, but they're very easy to build. I think we're at the infancy of this in video games. We're at the very beginning of it, and we're going to see way, way more of it. It doesn't necessarily have to be presenting it to your friends, or to an audience. I think the process of creating for a human being is fun in itself. MindsEye has been positioned as a linear game. You are best known for creating open world games. What was behind the decision to make MindsEye a more linear, narrative-driven experience? I think certain stories are more difficult to present to players in an open world setting. Open world gives you freedom – you don't necessarily want freedom to portray a story. For MindsEye, it's a very set time in a character, Jacob Diaz's, life. You pick up as Jacob when he arrives in Redrock, and then you leave Jacob at a certain point in the future. And so, it'd be very difficult for us to have an open world in there. It's horses for courses: it depends what you're doing. But for Jacob's story, it had to be a linear game. Having said that, there are open world experiences in there, and we can build them through Build.MindsEye. There is a free roam open world mode, where you play [as] a different character and you see his time, from the end of MindsEye, to the point of our next big planned launch. Again, they're all connected through a narrative, and we really want to show the universe, show the stories that have taken place in the universe, the characters in that universe, and see how they've experienced the same experience but from different viewpoints. "The dream from the building side is to allow players the opportunity to create their own multiplayer open world games with ease" Leslie Benzies, Build A Rocket Boy Was there ever a discussion about creating a more traditional GTA competitor? In design, you look at a lot of different options. I'm not sure it would've been smart as a company to say, 'we are going to compete with the biggest game on the planet'. I'm not sure that would be the best business decision to make. We went through a bunch of different designs, and to tell our story, this is what we landed on. MindsEye is priced more like a game from a decade ago at $60, and it'll take around 20 hours to finish. Can you talk about how you settled on the game's length and scope, and how you made that decision around price? So you've got the MindsEye campaign, and yes, it'll be about 20ish hours. But you do have all this other side content: there's going to be this continuous stream of content. These days, there are so many different options for people. It's not just games: there's streaming TV, so many good shows out there. I don't think you can have filler content in games. I think people want the meat, and they want the potatoes. We've tried to make as much meat as we can, if that makes sense. I think that's a good length for a game. What you also find through data, is that [with] big games, people don't play them all. The majority of people – 60% or 70% of people – don't actually play games to the end. So when you're making something, I would prefer – I'm sure the team would say the same – [that] you had the whole experience from start to finish, and not create this 200-hour game. Create something that is finishable, but have some side things that will fill out the universe. A lot of the side missions on the play side of MindsEye do fill out the characters' back stories, or do fill out what was happening in the world. On price: the world's in a funny place. People are worried about the price of eggs. So value for money, I think people appreciate that when times are difficult. I was curious why you waited until quite late in the day to reveal the build element of the game, only because it seemed you were being quite church and state with how MindsEye is releasing versus what Everywhere is. So in general, we believe – and again, it goes back to the amount of information, the amount of options people have these days – I don't think you can have extended marketing times. It's very expensive, we're a start-up. I think you lose interest from people. There are so many things for people to do, that if you extend it, you're not punching through to the place you need to be. I've seen other games, nine years before launch, it's getting talked about. I'm not sure that's the way of the world these days. You'll see there are games that never go to market: the day of launch was the marketing campaign, and it worked very well. So I think we tried to compress ours down for that reason. On the MindsEye.Play [continuous content] part of it, yeah, maybe we should've got that out there sooner, but it is a nice little surprise to give players. That's the thing with marketing – you never know what's the right or wrong way to do it, you've got to go with your gut, your senses, and test it. Being who you are, it brings a certain level of expectation and attention. Do you find it a double-edged sword, launching a new studio and launching a new game, with your background? Yes. There's always comparisons, and I think that's how humans work. As kids, we're taught to put a triangle into a triangular hole, and a square into a square hole. I think we do that for the rest of our lives, and we like to describe something new as 'it's X plus Y, with a bit of Z in there'. It makes things easy for us. It's maybe humans optimising the way we communicate. So there are comparisons. It serves us well in some ways, it doesn't serve us well in others. Dave Grohl said it well when he formed the Foo Fighters: nobody's interested in the Foo Fighters, all they were interested in was Nirvana. The guys have built something very cool, and I just hope people can see it for what it's trying to be.

This article contains spoilers for Fear Street: Prom Queen Right before the first kill of Fear Street: Prom Queen, viewers get a montage of sounds and images appropriate to the movie’s 1988 setting. Accompanied by Billy Idol’s “White Wedding,” we see mean girl Tiffany Falconer (Fina Strazza) prepping for prom, cool girl Christy Renault (Ariana Greenblatt) buying drugs, and protagonist Lori Granger (India Fowler) trying on her dress. Included in this sequence is a shot of moviegoers watching Phantasm II, the cult classic from director Don Coscarelli. Onscreen, we see one of that ’80s film’s standout scenes, where flying silver balls launched by the evil Tall Man (Angus Scrimm) soar across a hallway before lodging into the head of a victim. By this point, such homages are an expected part of the Fear Street franchise, based on the YA novels by R.L. Stine. The first three movies, all directed by Leigh Janiak, feature multiple references to previous horror movies. But the Phantasm II nod in Prom Queen does more than provide further watch suggestions for Fear Street‘s intended audience, young teens new to the horror genre. It also points to the important fundamentals of horror filmmaking, which is something sometimes lost in the more frantic original trilogy. As in Coscarelli’s movie, Prom Queen uses a deliberate pace and clear spacial arrangement to make its kills legible, thereby heightening the experience beyond jump scares and references. It makes for a better horror experience. Two Types of Killing Palmer’s approach can be seen in the scene directly following the montage. The scene opens with a mid-range shot of Greenblatt’s cool girl standing next to the sports car driven by her drug dealers. The camera stays pretty tight on her as she finishes her transaction and walks off. When she crosses a bridge to the parking lot where she’ll meet her end though, the visual language changes. A wide shot captures Christy as she crosses a bridge, the soft orange light from the post above her providing a tranquil mood. That continues into the next shot where our victim walks happily through an empty parking lot. The tone shifts suddenly with a hard cut to a close up of Christy as she stops, and the camera moves around her as she looks around. A cut back to a wide shot shows that she is largely alone and trapped on three sides. Brick buildings and trucks surround her on all corners. So when a figure arrives in a foreground of one of the wide shots and an axe drops into frame, we understand Christy’s predicament. She’s utterly alone, save for this person who’s come to kill her. And the killer stands in front of her one means of escape. Christy doesn’t see the killer before the axe blade lands in her neck, but we viewers do. In fact, the camera stays with the killer up until the attack, cutting only briefly back to Christy’s perspective. Even as the killer chops at the pleading girl, the camera returns to a brief wide shot to show how her cries will go unanswered. Contrast this kill to the opening scene of the first entry, Fear Street: 1994, in which a slasher in a mask and hood stalks bookseller Heather (Maya Hawke). Director Janiak uses the aisles of the mall storefronts where the attack takes place to to build tension. We never really know where the killer is in relationship to Heather, even when closeups show the characters in full. Thus the attacks are all shocks, with the killer (and in one case, a counterattacking Heather) coming out of nowhere to surprise the viewer. There’s nothing inherently wrong with this style. Janiak is a good filmmaker, and her excellent debut picture Honeymoon demonstrates that she knows how to build dread and mood. But with Fear Street, Janiak chooses shocks over scares, with the killers constantly jumping out at the victims. Such energetic moviemaking gets a jolt from viewers but ultimately operates as shorthand, giving us the quick payoff of a scary scene without making us wait through the buildup. Prom Queen is all about the fundamentals of building a scary scene. In the bloody art room sequence, perhaps the most memorable moment, the camera lays out the geography of the space before we see the killer use a paper cutter to sever a jock’s arms. A tracking shot follows the jock Bobby (Dakota Taylor) and his girlfriend Linda (Ilan O’Driscoll) as they enter the room, keeping them in the center of the frame while also showing the points of exit and the objects in the room. Linda gets killed off-screen, moving out of frame and into the darkness. When she returns, holding the entrails spilling from a wound in her stomach, we see not only the room, but also start looking for the instruments that can do such damage. So when the killer stymies Bobby’s counterattack by slamming his arm on a table, we’re not seeing the paper cutter for the first time. But we do feel a sense of dread when we realize the proximity between Bobby’s hand and the cutter, a dread that gets consummated when he starts chopping away. Join our mailing list Get the best of Den of Geek delivered right to your inbox! It’s that sense of dread that makes horror so powerful. The actual kill can surprise us or disgust us, and Fear Street does that well (remember the improbable head in a bread slicer kill from 1994?). But the kill can also provide a sense of relief because it’s all so knowable. We no longer fear the potential for a character’s death because it has arrived. We may hate it and it may upset us, but it doesn’t scare us. A Legacy of Death With its more intentional pace and excellent spacial awareness, Prom Queen spends more time building dread. Drawing from masters like Coscarelli (who certainly loved a good jump scare from time to time) and pulling from the Italian giallo tradition that inspired the American slasher, director Matt Palmer grounds Prom Queen in a larger horror tradition. Palmer shows his young audience not just the cool movies they can check out after watching the latest Fear Street. He also shows them how to watch these movies; the fundamentals of horror filmmaking that provide the building blocks of the genre. Thanks to its well-constructed and well-paced kill scenes, Prom Queen goes beyond referencing the 1980s and looks at the history of horror and educates the next generation of horror fans. Fear Street: Prom Queen streams on Netlfix on May 23, 2025.

May 23, 20257 min readWhat FDA’s Planned Limits on COVID Vaccinations Mean for HealthDespite the fact that vaccines against COVID have already undergone strict safety reviews and that people continue to die from the disease, Trump’s FDA is moving to reduce access to annual COVID boosters for healthy AmericansBy Stephanie Armour & KFF Health News aire images/Getty ImagesLarry Saltzman has blood cancer. He’s also a retired doctor, so he knows getting covid-19 could be dangerous for him — his underlying illness puts him at high risk of serious complications and death. To avoid getting sick, he stays away from large gatherings, and he’s comforted knowing healthy people who get boosters protect him by reducing his exposure to the virus.Until now, that is.Vaccine opponents and skeptics in charge of federal health agencies — starting at the top with Health and Human Services Secretary Robert F. Kennedy Jr. — are restricting access to covid shots that were a signature accomplishment of President Donald Trump’s first term and cost taxpayers about $13 billion to develop, produce, and distribute. The agencies are narrowing vaccination recommendations, pushing drugmakers to perform costly clinical studies, and taking other steps that will result in fewer people getting protection from a virus that still kills hundreds each week in the U.S.On supporting science journalismIf you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.“There are hundreds of thousands of people who rely on these vaccines,” said Saltzman, 71, of Sacramento, California. “For people who are immunocompromised, if there aren’t enough people vaccinated, we lose the ring that’s protecting us. We’re totally vulnerable.”The Trump administration on May 20 rolled out tougher approval requirements for covid shots, described as a covid-19 “vaccination regulatory framework,” that could leave millions of Americans who want boosters unable to get them.The FDA will encourage new clinical trials on the widely used vaccines before approving them for children and healthy adults. The requirements could cost drugmakers tens of millions of dollars and are likely to leave boosters largely out of reach for hundreds of millions of Americans this fall.Under the new guidance, vaccines will be available for high-risk individuals and seniors. But the FDA will encourage drugmakers to commit to conducting post-marketing clinical trials in healthy adults when the agency approves covid vaccines for those populations.For the past five years, the shots have been recommended by the Centers for Disease Control and Prevention for everyone 6 months and older. They have been available each fall after being updated to reflect circulating strains of the virus, and the vaccines have been shown to be safe and effective in clinical trials.Vinay Prasad, who leads the FDA’s division overseeing vaccines, cited “distrust of the American public” as he announced the new guidelines at a May 20 briefing.“We have launched down this multiyear campaign of booster after booster after booster,” he said, adding that “we do not have gold-standard science to support this for average-risk, low-risk Americans.”The details were outlined in a May 20 article in The New England Journal of Medicine, written by FDA Commissioner Marty Makary. He and Prasad later followed up with the briefing, which appeared the same day on YouTube.The added limits on access aren’t the result of any recent data showing there are new health risks from the covid vaccines. Instead, they reflect a different regulatory stance from Kennedy, who has a history of anti-vaccine activism, and Makary, who has questioned the safety data on covid mRNA shots.Announcing a major regulatory change in a medical journal and YouTube video is a highly unusual approach that still leaves many questions about implementation unanswered. It remains unclear when the changes will go into effect or whether there will be any public comment period. The changes were announced by the administration before an FDA advisory committee meeting on May 22 to consider the 2026 covid vaccine formula.It’s a sharp reversal from the first Trump administration, which launched Operation Warp Speed — the effort that led to the development of the covid shots. Trump called the vaccines the “gold standard” and a “monumental national achievement.”Concerns About Higher TransmissionThe announcement is rattling some patient advocacy groups, doctors, nursing home leaders, and researchers who worry about the ramifications. They say higher-risk individuals will be more likely to get covid if people who aren’t at risk don’t get boosters that can help reduce transmission. And they say the FDA’s restrictions go too far, because they don’t provide exceptions for healthy individuals who work in high-risk settings, such as hospitals, who may want a covid booster for protection.The limits will also make it harder to get insurance coverage for the vaccines. And the FDA’s new stance could also increase vaccine hesitancy by undermining confidence in covid vaccines that have already been subject to rigorous safety review, said Kate Broderick, chief innovation officer at Maravai Life Sciences, which makes mRNA products for use in vaccine development.“For the public, it raises questions,” she said. “If someone has concerns, I’d like them to know that of all the vaccines, the ones with the most understood safety profile are probably covid-19 vaccines. There is an incredible body of data and over 10 billion doses given.”Some doctors and epidemiologists say it could leave healthy people especially vulnerable if more virulent strains of covid emerge and they can’t access covid shots.“It’s not based on science,” said Rob Davidson, an emergency room doctor in Michigan and executive director of the Committee to Protect Health Care, which works to expand health care access. “It’s what we were all worried would happen. It risks peoples’ lives.”Current federal regulators say there is no high-quality evidence showing that vaccinating healthy people, including health workers who are near or around immunocompromised people, provides an additional benefit.“It is possible, actually, that such approvals and strategies provide false reassurance and lead to increased harms,” Prasad said.The covid vaccines underwent clinical trials to assess safety, and they have been subject to ongoing surveillance and monitoring since they obtained emergency use authorization from the FDA amid the pandemic. Heart issues and allergic reactions can occur but are rare, according to the CDC.On a separate track, the FDA on May 21 posted letters sent in April to makers of the mRNA covid vaccines to add information about possible heart injury on warning labels, a move that one former agency official described as overkill. The action came after the Permanent Subcommittee on Investigations, a panel of the Senate Homeland Security and Governmental Affairs Committee, held a hearing on alleged adverse events associated with covid vaccines.Limiting boosters to healthy people goes against guidance from some medical groups.“The COVID-19 vaccine is safe, effective, and the best way to protect children,” Sean O’Leary, chair of the Committee on Infectious Diseases at the American Academy of Pediatrics, said in an email. “Young children under 5 continue to be at the highest risk, with that risk decreasing as they get older.”Unsupported Claims About mRNA VaccinesThe covid booster clampdown is supported by many adherents of the “Make America Healthy Again” movement, which casts suspicion on traditional medicine. Some opponents of covid mRNA vaccines say without evidence that the shots cause “turbo” cancer, are genetic bioweapons, and cause more heart damage than the covid virus.There is no evidence the shots lead to rapid and aggressive cancers. Cancer rates decreased an average of 1.7% per year for men and 1.3% for women from 2018 to 2022, according to the National Institutes of Health. The covid vaccines debuted in 2021.Federal regulators say narrowing who can get the boosters will align the U.S. with policies of European nations. But other countries have vastly different economic structures for health care and approaches to preventive care. Many European countries, for example, don’t recommend flu shots for the entire population. The U.S. does in part because of the financial drain attributed to lost productivity when people are sick.They also want more information. “I think there’s a void of data,” Makary told CBS News on April 29. “And I think rather than allow that void to be filled with opinions, I’d like to see some good data.”A massive five-year study on covid vaccine safety by the Global Vaccine Data Network, involving millions of people, was underway, with about a year left before completion. The Trump administration terminated funding for the project as part of cuts directed by the president’s Department of Government Efficiency, and work on the study has stopped for now.There are a multitude of studies, however, on the vaccines’ effectiveness in preventing severe illness, hospitalization, and death.Limiting boosters for healthy people can be risky, some doctors say, because people don’t always know when they fall into higher-risk categories, such as individuals who are prediabetic or have high blood pressure. The covid vaccine restrictions could deter them from getting boosted, and they might experience worse complications from the virus as a result. For example, about 40% of people with hepatitis C are unaware of their condition, according to a study published in 2023.The number of people getting covid vaccines has already dropped significantly since the height of the crisis. More than half of the more than 258 million adults in the U.S. had gotten a covid vaccination as of May 2021, according to the CDC. In each of the past two seasons, less than 25% of Americans received boosters, CDC data shows.While deaths from the virus have dropped, covid remains a risk, especially when cases peak in December and January. Weekly covid deaths topped 2,580 as recently as January 2024, according to CDC data.Some high-risk individuals are worried that the new restrictions are just the first salvo in halting all access to mRNA shots. “The HHS motivation really is hidden, and it’s to dismiss all mRNA technology,” said Michael Osterholm, an epidemiologist at the University of Minnesota.Officials at the NIH have told scientists to remove references to mRNA in grant applications. HHS also announced plans in May to develop new vaccines without mRNA technology, which uses messenger RNA to instruct cells to make proteins that trigger an immune response.Rose Keller, 23, is concerned about future access to covid shots. She would be eligible under the current announcement — she has cystic fibrosis, a progressive genetic condition that makes the mucus in her lungs thick and sticky, so covid could land her in the hospital. But she is concerned the Trump administration may go further and restrict access to the vaccines as part of a broader opposition to mRNA technology.“I’ve had every booster that’s available to me,” said Keller, a government employee in Augusta, Maine. “It’s a real worry if I don’t have the protection of a covid booster.”KFF Health News is a national newsroom that produces in-depth journalism about health issues and is one of the core operating programs at KFF — the independent source for health policy research, polling, and journalism.

Tech AI to monitor NYC subway safety as crime concerns rise AI-powered surveillance aims to prevent crime before it happens in Big Apple’s subway system Published May 23, 2025 6:00am EDT close 'Decoding Broken Hearts': AI used to advance heart health research Fox News anchor Bret Baier has the latest on the Murdoch Children's Research Institute's partnership with the Gladstone Institutes for the "Decoding Broken Hearts" initiative on "Special Report." Imagine having a tireless guardian watching over you during your subway commute. It notices every detail with steady focus, all with one goal in mind: keeping you safe on the subway.New York City's subway system is testing artificial intelligence to boost security and reduce crime. Michael Kemper, a 33-year NYPD veteran and the chief security officer for the Metropolitan Transportation Authority (MTA), which is the largest transit agency in the United States, is leading the rollout of AI software designed to spot suspicious behavior as it happens. The MTA says this technology represents the future of subway surveillance and reassures riders that privacy concerns are being taken seriously.JOIN THE FREE ‘CYBERGUY REPORT’: GET MY EXPERT TECH TIPS, CRITICAL SECURITY ALERTS AND EXCLUSIVE DEALS, PLUS INSTANT ACCESS TO MY FREE ‘ULTIMATE SCAM SURVIVAL GUIDE’ WHEN YOU SIGN UP! New York City subway (Kurt "CyberGuy" Knutsson)WHAT IS ARTIFICIAL INTELLIGENCE (AI)?How AI surveillance will work in the New York City SubwayThe AI system will analyze real-time camera feeds across subway platforms and train cars to detect unusual or potentially dangerous behaviors. Instead of identifying individuals, the technology focuses on behavior patterns that might signal trouble, such as erratic movements or confrontations. When the AI flags something suspicious, it can alert transit police to respond quickly, which could help prevent crimes before they escalate.Currently, about 40% of subway cameras are monitored live by staff. The AI aims to expand this coverage without needing additional personnel, which could enhance safety across the system. While details about the AI providers and full deployment timelines have not been shared, the MTA emphasizes that facial recognition technology will not be part of this system. This is intended to protect rider privacy. A surveillance camera mounted on a building (Kurt "CyberGuy" Knutsson)Crime trends and government responseCrime in NYC subways has seen ups and downs over recent years. While major crimes dropped by about 7.7% through mid-2024, assaults have unfortunately increased. To bolster safety, New York Gov. Kathy Hochul deployed 1,000 National Guard members to support transit police efforts. AI technology is viewed as a complementary tool in this broader strategy to keep subways safe, not a standalone solution.'DELUSIONAL' NYC POLITICIANS CALLED OUT BY GUARDIAN ANGELS FOUNDER OVER SUBWAY VIOLENCE: 'SLAP IN THE FACE' A police vehicle (Kurt "CyberGuy" Knutsson)Is AI being used in public transit elsewhere?New York is not alone in exploring AI for transit security. Cities like Los Angeles and Chicago have experimented with AI-powered cameras for tasks ranging from weapons detection to issuing parking citations. These initiatives reflect a growing interest in using technology to enhance public safety in transit environments. Entrance to NYC subway (Kurt "CyberGuy" Knutsson)Kurt's key takeawaysWhile the promise of AI surveillance is encouraging, several questions remain unanswered. We do not yet know exactly which behaviors the system will flag or how effective it will be in practice. Perhaps more importantly, privacy advocates and civil liberties groups warn about potential biases in AI systems and the risk of false positives that could unfairly target certain groups.The MTA's decision to exclude facial recognition is a positive step, but ongoing transparency and oversight will be crucial. Michael Kemper's extensive experience leading the NYPD's Transit Bureau, where he helped reduce subway crime, adds credibility to the initiative. Still, as this technology rolls out, balancing safety with privacy and fairness will be key to earning public trust.CLICK HERE TO GET THE FOX NEWS APPWould you trust AI to watch over you in the subway or are you more concerned about your privacy? Let us know by writing us atCyberguy.com/Contact.For more of my tech tips and security alerts, subscribe to my free CyberGuy Report Newsletter by heading to Cyberguy.com/Newsletter.Ask Kurt a question or let us know what stories you'd like us to cover.Follow Kurt on his social channels:Answers to the most-asked CyberGuy questions:New from Kurt:Copyright 2025 CyberGuy.com. All rights reserved. Kurt "CyberGuy" Knutsson is an award-winning tech journalist who has a deep love of technology, gear and gadgets that make life better with his contributions for Fox News & FOX Business beginning mornings on "FOX & Friends." Got a tech question? Get Kurt’s free CyberGuy Newsletter, share your voice, a story idea or comment at CyberGuy.com.

May 22, 20253 min readHow Much Ultraprocessed Food Do You Eat? Blood and Urine Record ItA new study suggests blood and urine samples could provide an objective measure of diets and help unravel their connections to diseaseBy Smriti Mallapaty & Nature magazine Diets rich in industrially manufactured food have been associated with several health conditions. monticelllo/Getty ImagesMolecules in urine and blood can reveal how much of a person’s diet comes from ultra-processed foods, according to a study published in PLOS Medicine today. The paper suggests that these measurements provide an objective way to track consumption of ultra-processed food — and would be useful for investigating links to diseases such as diabetes and cancer.Ultra-processed foods are industrially manufactured and often contain ingredients, such as additives and emulsifiers, that are not typically found in home-cooked food. From sweetened yogurts to factory-made bread and packaged snacks, “it’s a really wide range of foods”, says study co-author Erikka Loftfield, an epidemiologist at the US National Cancer Institute in Rockville, Maryland.Studies have linked high consumption of ultra-processed food to an increased risk of obesity, cardiovascular disease, diabetes and some types of cancer. But these investigations have typically relied on asking individuals to remember what they have eaten, which can be unreliable.On supporting science journalismIf you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.Studies have also shown the potential of studying the products of metabolism excreted in the blood and urine, but only for a small number of such metabolites. Loftfield and her colleagues have now expanded that analysis to include more than 1,000 metabolites, which are produced when the body converts food into energy.“This work is important,” says Oliver Robinson, a molecular epidemiologist at Imperial College London. “There’s a lot of measurement error in traditional assessment methods for diet.”Diet scoringLoftfield and her colleagues studied samples collected from 718 healthy individuals aged 50–74 in 2012–13. The participants’ urine and blood samples had been collected twice, six months apart. The participants had also been asked — up to six times during the one-year study period — to record everything they had consumed the day before. Loftfield’s team labelled each food item, from the bun in a burger to the cheese slice, meat patty and ketchup, either ultra-processed or not.The researchers then used a machine-learning technique to give each participant a score for how much of their daily energy intake came from ultra-processed foods, says Loftfield. They found that, on average, ultra-processed food accounted for 50% of the participants’ energy intake, but the range varied between individuals — from 12% to 82%. Individuals who consumed the most ultra-processed foods generally got more of their energy from carbohydrates, added sugars and saturated fats — and less from proteins and fibre — than did those who consumed lower levels of ultra-processed foods.The researchers then analysed the participants’ blood and urine samples, looking for metabolites that were more common in the people whose diets contained lots of ultra-processed food.Samples from people with diets rich in ultra-processed foods were more likely to contain a metabolite linked to an increased risk of type-2 diabetes — and some of these people’s urine samples contained a molecule produced by certain food packaging. They also contained fewer metabolites derived from fresh fruits and vegetables.Randomized studyTo test whether the metabolite scoring could be used to predict the presence of a lot of ultra-processed food in a person’s diet, the researchers used data from a randomized controlled diet study of 20 individuals aged 18–50 conducted between 2018 and 2020. Half of the participants were asked to eat a diet rich in ultra-processed food for two weeks, then a diet containing no ultra-processed food for two weeks; the rest ate the same diets in the opposite order. Loftfield and her colleagues found they were able to distinguish between the two diets by looking at the metabolites in the participants’ urine and blood samples.Loftfield wants to test the method on populations with more varied diets, and on younger people, who tend to eat more ultra-processed foods. She hopes to use the tool to improve studies of the link between ultra-processed food and cancer.Robinson wonders whether the tool could be used to address big unanswered questions, including what it is about ultra-processed food that is bad for you. It is difficult to distinguish between the metabolic signatures of a diet rich in ultra-processed food and of one that is unhealthy because it is high in salt, sugar and saturated fats, he says.Understanding this difference could help companies to improve their products. “We’re sort of trapped in this industrial food-production system where we all eat ultra-processed food, and it’s quite hard for most people to go back to fresh food, because our food systems are now set up like this,” adds Robinson.This article is reproduced with permission and was first published on May 20, 2025.

It's shocking! Scientists figure out how the brain forms emotional connections Neural recordings track how neurons link environments to emotional events. Jacek Krywko – May 21, 2025 4:07 pm | 16 Credit: fotografixx Credit: fotografixx Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only   Learn more Whenever something bad happens to us, brain systems responsible for mediating emotions kick in to prevent it from happening again. When we get stung by a wasp, the association between pain and wasps is encoded in the region of the brain called the amygdala, which connects simple stimuli with basic emotions. But the brain does more than simple associations; it also encodes lots of other stimuli that are less directly connected with the harmful event—things like the place where we got stung or the wasps’ nest in a nearby tree. These are combined into complex emotional models of potentially threatening circumstances. Till now, we didn’t know exactly how these models are built. But we’re beginning to understand how it’s done. Emotional complexity “Decades of work has revealed how simple forms of emotional learning occurs—how sensory stimuli are paired with aversive events,” says Joshua Johansen, a team director at the Neural Circuitry of Learning and Memory at RIKEN Center for Brain Science in Tokyo. But Johansen says that these decades didn’t bring much progress in treating psychiatric conditions like anxiety and trauma-related disorders. “We thought if we could get a handle of more complex emotional processes and understand their mechanisms, we may be able to provide relief for patients with conditions like that,” Johansen claims. To make it happen, his team performed experiments designed to trigger complex emotional processes in rats while closely monitoring their brains. Johansen and Xiaowei Gu, his co-author and colleague at RIKEN, started by dividing the rats into two groups. The first “paired” group of rats was conditioned to associate an image with a sound. The second “unpaired” group watched the same image and listened to the same sound, but not at the same time. This prevented the rats from making an association. Then, one day later, the rats were shown the same image and treated with an electric shock until they learned to connect the image with pain. Finally, the team tested if the rats would freeze in fear in response to the sound. The “unpaired” group didn’t. The rats in the “paired” group did—it turned out human-like complex emotional models were present in rats as well. Once Johansen and Gu confirmed the capacity was there, they got busy figuring out how it worked exactly. Playing tag “Behaviorally, we measured freezing responses to the directly paired stimulus, which was the image, and inferred stimulus which was the sound,” Johansen says. “But we also performed something we called miniscope calcium imaging.” The trick relied on injecting rats with a virus that forced their cells to produce proteins that fluoresce in response to increased levels of calcium in the cells. Increased levels of calcium are the telltale sign of activity in neurons, meaning the team could see in real time which neurons in rats’ brains lit up during the experiments. It turned out that the region crucial for building these complex emotional models was not the amygdala, but the dorsomedial prefrontal cortex (dmPFC), which had a rather specialized role. “The dmPFC does not form the sensory model of the world. It only cares about things when they have emotional relevance,” Johansen explains. He said there wasn’t much change in neuronal activity during the sensory learning phase, when the animals were watching the image and listening to the sound. The neurons became significantly more active when the rats received the electric shock. In the “unpaired” group, the active neurons that held the representations of the electric shock and the image started to overlap. In the “paired” group, this overlap also included the neuronal representation of the sound. “There was a kind of an associative bundle that formed,” Johansen says. After Johansen and Gu pinpointed the neurons that formed those associative bundles, they started looking at how each of these components works. Detraumatizing rodents In the first step, the team identified the dmPFC neurons that sent output to the amygdala. Then they selectively inhibited those neurons and exposed the rats from the “paired” group to the image and the sound again. The result of disconnecting the dmPFC neurons from the amygdala was that rats exhibited a fear response to the image but no longer feared the sound. “It seems like the amygdala can form the simple representations on its own but requires input from the dmPFC to express more complex, inferred emotions,” Johansen says. But there are still a lot of unanswered questions left. The next thing the team wants to take a closer look at is the process that enables the brain to tie an aversive stimulus, like the shock, to one that was not active during the aversive event. In the “paired” group of rats, some multi-sensory neurons responding to both auditory and visual stimuli apparently got recruited. “We haven’t worked that out yet,” Johansen says. "This is a very novel type of mechanism.” Another thing is that the emotional model Johansen and Gu induced in rats was relatively simple. In the real world, especially in humans, we can have many different aversive outcomes tied to the same triggers. A single location could be where you got stung by a wasp, attacked by a dog, robbed of your wallet, and dumped by your significant other—all different aversive representations with myriad inferred, indirect stimuli to go along with them. “Does the dmPFC combine all those representations into sort of a single, overlapping representation? Or is it a really rich environment that bundles different aversive experiences with the individual aspects of these experiences?” Johansen asked. “This is something we want to test more.” Nature, 2025.  DOI: 10.1038/s41586-025-09001-2 Jacek Krywko Associate Writer Jacek Krywko Associate Writer Jacek Krywko is a freelance science and technology writer who covers space exploration, artificial intelligence research, computer science, and all sorts of engineering wizardry. 16 Comments

This is such exciting work about environmental forecasting, so we’re happy to have the two of you join us today.   Megan and Wessel, welcome.  MEGAN STANLEY: Thank you. Thanks. Great to be here.  WESSEL BRUINSMA: Thanks.  TINGLE: Let’s jump right in. Wessel, share a bit about the problem your research addresses and why this work is so important.  BRUINSMA: I think we’re all very much aware of the revolution that’s happening in the space of large language models, which have just become so strong. What’s perhaps lesser well-known is that machine learning models have also started to revolutionize this field of weather prediction. Whereas traditional weather prediction models, based on physical laws, used to be the state of the art, these traditional models are now challenged and often even outperformed by AI models.   This advancement is super impressive and really a big deal. Mostly because AI weather forecasting models are computationally much more efficient and can even be more accurate. What’s unfortunate though, about this big step forward, is that these developments are mostly limited to the setting of weather forecasting.   Weather forecasting is very important, obviously, but there are many other important environmental forecasting problems out there, such as air pollution forecasting or ocean wave forecasting. We have developed a model, named Aurora, which really kicks the AI revolution in weather forecasting into the next gear by extending these advancements to other environmental forecasting fields, too. With Aurora, we’re now able to produce state-of-the-art air pollution forecasts using an AI approach. And that wasn’t possible before!  TINGLE: Megan, how does this approach differ from or build on work that’s already been done in the atmospheric sciences?  STANLEY: Current approaches have really focused training very specifically on weather forecasting models. And in contrast, with Aurora, what we’ve attempted to do is train a so-called foundation model for the Earth system. In the first step, we train Aurora on a vast body of Earth system data. This is our pretraining step.   And when I say a vast body of data, I really do mean a lot. And the purpose of this pretraining is to let Aurora, kind of, learn some general-purpose representation of the dynamics that govern the Earth system. But then once we’ve pretrained Aurora, and this really is the crux of this, the reason why we’re doing this project, is after the model has been pretrained, it can leverage this learned general-purpose representation and efficiently adapt to new tasks, new domains, new variables. And this is called fine-tuning.  The idea is that the model really uses the learned representation to perform this adaptation very efficiently, which basically means Aurora is a powerful, flexible model that can relatively cheaply be adapted to any environmental forecasting task.    TINGLE: Wessel, can you tell us about your methodology? How did you all conduct this research?  BRUINSMA: While approaches so far have trained models on primarily one particular data These datasets are a combination of estimates of the historical state of the world, forecasts by other models, climate simulations, and more. We’ve been able to show that training on not just more data but more diverse data helps the model achieve even better performance. Showing this is difficult because there is just so much data.   In addition to scaling to more and more diverse data, we also increased the size of the model as much as we could. Here we found that bigger models, despite being slower to run, make more efficient use of computational resources. It’s cheaper to train a good big model than a good small model. The mantra of this project was to really keep it simple and to scale to simultaneously very large and, more importantly, diverse data and large model size.  TINGLE: So, Megan, what were your major findings? And we know they’re major because they’re in Nature. [LAUGHS]  STANLEY: Yeah, [LAUGHS] I guess they really are. So the main outcome of this project is we were actually able to train a single foundation model that achieves state-of-the-art performance in four different domains. Air pollution forecasting. For example, predicting particulate matter near the surface or ozone in the atmosphere. Ocean wave forecasting, which is critical for planning shipping routes.   Tropical cyclone track forecasting, so that means being able to predict where a hurricane or a typhoon is expected to go, which is obviously incredibly important, and very high-resolution weather forecasting.   And I’ve, kind of, named these forecasting domains as if they’re just items in a list, but in every single one, Aurora really pushed the limits of what is possible with AI models. And we’re really proud of that.   But perhaps, kind of, you know, to my mind, the key takeaway here is that the foundation model approach actually works. So what we have shown is it’s possible to actually train some kind of general model, a foundation model, and then adapt it to a wide variety of environmental tasks. Now we definitely do not claim that Aurora is some kind of ultimate environmental forecasting model. We are sure that the model and the pretraining procedure can actually be improved. But, nevertheless, we’ve shown that this approach works for environmental forecasting. It really holds massive promise, and that’s incredibly cool.  TINGLE: Wessel, what do you think will be the real-world impact of this work?  BRUINSMA: Well, for applications that we mentioned, which are air pollution forecasting, ocean wave forecasting, tropical cyclone track forecasting, and very high-resolution weather forecasting, Aurora could today be deployed in real-time systems to produce near real-time forecasts. And, you know, in fact, it already is. You can view real-time weather forecasts by the high-resolution version of the model on the website of ECMWF (European Centre for Medium-Range Weather Forecasts).  But what’s remarkable is that every of these applications took a small team of engineers about four to eight weeks to fully execute. You should compare this to a typical development timeline for more traditional models, which can be on the order of multiple years. Using the pretraining fine-tuning approach that we used for Aurora, we might see significantly accelerated development cycles for environmental forecasting problems. And that’s exciting.  TINGLE: Megan, if our listeners only walk away from this conversation with one key talking point, what would you like that to be? What should we remember about this paper?  STANLEY: The biggest takeaway is that the pretraining fine-tuning paradigm, it really works for environmental forecasting, right? So you can train a foundational model, it learns some kind of general-purpose representation of the Earth system dynamics, and this representation boosts performance in a wide variety of forecasting tasks. But we really want to emphasize that Aurora only scratches the surface of what’s actually possible.  So there are many more applications to explore than the four we’ve mentioned. And undoubtedly, the model and pretraining procedure can actually be improved. So we’re really excited to see what the next few years will bring.  TINGLE: Wessel, tell us more about those opportunities and unanswered questions. What’s next on the research agenda in environmental prediction?  BRUINSMA: Well, Aurora has two main limitations. The first is that the model produces only deterministic predictions, by which I mean a single predicted value. For variables like temperature, this is mostly fine. But other variables like precipitation, they are inherently some kind of stochastic. For these variables, we really want to assign probabilities to different levels of precipitation rather than predicting only a single value.  An extension of Aurora to allow this sort of prediction would be a great next step.   The second limitation is that Aurora depends on a procedure called assimilation. Assimilation attempts to create a starting point for the model from real-world observations, such as from weather stations and satellites. The model then takes the starting point and uses it to make predictions. Unfortunately, assimilation is super expensive, so it would be great if we could somehow circumvent the need for it.  Finally, what we find really important is to make our advancements available to the community. [MUSIC]  TINGLE: Great. Megan and Wessel, thanks for joining us today on the Microsoft Research Podcast.  BRUINSMA: Thanks for having us.  STANLEY: Yeah, thank you. It’s been great.  TINGLE: You can check out the Aurora model on Azure AI Foundry. You can read the entire paper, “A Foundation Model for the Earth System,” at aka.ms/abstracts. And you’ll certainly find it on the Nature website, too.   Thank you so much for tuning in to Abstracts today. Until next time.   [MUSIC FADES]