pastoral nomads, please stop here Archaeologists just mapped a Bronze Age megafortress in Georgia This recently mapped Bronze Age fortress is just one among hundreds. Kiona N. Smith Jan 9, 2025 7:00 am | 1 This orthographic photo shows the inner fortress walls and some nearby structures. Credit: Erb-Satullo et al. 2025 This orthographic photo shows the inner fortress walls and some nearby structures. Credit: Erb-Satullo et al. 2025 Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreA sprawling 3,500-year-old fortress offers tantalizing clues about a culture that once dotted the southern Caucasus mountains with similar walled communities.Archaeologists recently used a drone to map a sprawling 3,500-year-old fortress in the Caucasus Mountains of southern Georgia. The detailed aerial map offers some tantalizing clues about the ancient culture whose people built hundreds of similar fortresses in a mountainous region that spans the modern countries of Armenia, Azerbaijan, Georgia, and Turkey. Based on their survey and excavations within the fortress walls, Cranfield University archaeologist Nathaniel Erb-Satullo and his colleagues suggest the fortified community may have been a place where nomadic herders converged during their yearly migration, but the evidence still leaves more questions than answers. This map shows an aerial map of the ancient megafortress at Dmanisis Gora. Credit: Erb-Satullo et al. 2025 An abandoned ancient megafortressThe half-buried Bronze Age ruins of Dmanisis Gora perch on a windswept promontory a few kilometers away from a cave where Homo erectus (or a close relative) lived 1.8 million years ago. Deep, steep-sided gorges run along two sides of the promontory, and sometime between 1500 and 1000 BCE, people stacked boulders into a double layer of high, thick walls to block off the end of the plateau from the plains to the west. Sheltered between the 4-meter high, 2.5-meter wide walls and the 60-meter-deep gorges, people built dugout houses, then later aboveground stone ones, along with stone animal pens and other buildings.Outside the walls lay a more sprawling, less densely packed settlement, sheltered by another wall to the west. That outer wall was as high and wide as the inner ones, and it stretched a full kilometer from the edge of one gorge to the edge of the other. Between the walls, homes and other buildings formed small compounds with open space between them. Fenced fields, animal pens, and graves dotted the area.Erb-Satullo and his colleagues recently mapped the whole fortress with their DJI Phantom 4 RTK drone. Eleven-thousand aerial images, stitched together with software into orthographic photos and digital elevation maps, revealed that the fortified community was much larger than the team had initially suspected. The innermost walls shelter an area of about 1.5 hectares, but the other wall encloses a total of 56. And a partial wall, which may have been destroyed in the past or simply started and never finished, would bring the total fortified area to around 80 hectares.Because of its size, it was impossible to get a sense of the site as a whole from the ground, Erb-Satullo and his colleagues explain in a press release. This aerial photo shows the outer fortress wall, with the outlines of several collapsed structures in the foreground, looking west. Credit: Erb-Satullo et al. 2025 A world of fortresses and nomadsAlthough the fortress is hugeseveral times than the area of the nearby medieval town of Dmanisi, a major trading post during the Middle Ages that once boasted a cathedral and a castleits just one of hundreds that dot the southern Caucasus mountains, most of which havent been surveyed or excavated in detail.These examples suggest that the Damanisis Gora mega-fortress, while exceptional in size, is not entirely without parallel, write Erb-Satullo and his colleagues in their paper.These fortresses tended to take advantage of features like gorges and hilltops for natural defenses, supplementing those with walls of unworked boulders stacked without mortar. Within the walls, crafters cast elaborate bronze work and made gray and black pottery burnished to a shine (though reds and buffs got more popular over time). They sewed using bone needles, and they wore beads of bone, carnelian, copper, and faience (a type of ceramic).The people who lived and died in these walled communities buried their dead in mounds called kurgans, in massive stone-built tombs called cromlechs, or smaller stone-lined graves called cists (pronounced kist). One of these cist graves lies near the main gate of the inner fortress, positioned such that anyone passing through the gate would have passed directly by it. Its occupant went to the next life with beads, copper-alloy arrowheads, and pottery. In the inner and outer fortress, a mix of all three types of graves dot the area; theres no defined graveyard to separate the dwellings of the dead from those of the living.Smaller, narrower stone walls mark the boundaries of fields, both within and outside the walls of the fortified community. Some of those fields were plowed and fenced long after the fortress had crumbled into rubble; aerial photos reveal where plows tore into the foundations of ancient structures. But others may date to the Bronze Age and early Iron Age heyday of Dmanisis Gora.A settlement of migrants?For a large part of the community, however, making a living meant herding livestock, a life of constant migration between highland pastures in the summer and winter grazing in the lowlands.Erb-Satullo and his colleagues are still trying to understand how the constant seasonal motion of pastoral herders relates to settled life behind stone walls. But at the moment, their best idea is that the area between the inner and outer walls may have been a seasonal settlement for herders passing through. Dmanisis Gora lies on the route ancient herders would have followed each spring and autumn.One possibility is that Dmanisis Gora served as a staging ground for pastoral groups during transitional periods in the spring and autumn, write Erb-Satullo and his colleagues. In an earlier paper, following some excavations at the site in 2022, they suggested, Its large outer settlement may have expanded and contracted seasonally.A seasonal stopover for migrating herders and their flocks may explain some puzzling quirks of the outer fortress. People went to the trouble of building stone buildings and pens, which suggests a permanent settlement rather than a temporary camp, but they didnt leave behind many artifacts. In the innermost section of the fortress, archaeologists have unearthed two distinct layers of construction and discarded objects: tens of thousands of potsherds, beads, needles, and ritual items. But in the area between the innermost and outer walls, finds are sparse. That may mean that people stayed in the outer fortress regularly enough to justify stone buildings, but they probably didnt stick around for very long.That gave rise to the idea that the outer area may have been seasonal quarters for nomads, who took shelter behind the walls with their herds during the twice-yearly trek from lowland to highlands back again. To be sure, Erb-Satullo and his colleagues will need to study Dmanisis Gora more closely and make detailed maps of similar sites.With the site now extensively mapped, further study will start to provide insights into areas such as population density and intensity, livestock movements, and agricultural practices, among others, says Erb-Satullo in a press release. This orthographic photo shows the inner fortress walls and some nearby structures. Credit: Erb-Satullo et al. 2025 A tale of resilience and ruin?Another question on Erb-Satullo and his colleagues minds is how settlements like Dmanisis Gora weathered the Bronze Age collapse: a wave of invasions, famines, earthquakes, and economic and political upheaval that wreaked havoc on civilizations in Mesopotamia, the Nile Valley, and along the Mediterranean at the beginning of the 1100s BCE.At Dmanisis Gora, radiocarbon dates and the types of pottery and construction suggest that life carried on without a hitch (or at least not one that shows up in the archaeological record so far) even as the rest of the world was violently transitioning from the Bronze Age to the Iron Age. Erb-Satullo and his colleagues suggest that something about the blend of mobility and fortified settlement may have given the fortress-building people of the Caucasus some resilience against the tides of 12th-century collapse. But thats another question that will ultimately require more evidence to answer.Meanwhile, reminding us that no civilizationno matter how resilientis actually permanent, the aerial photos reveal where plows had torn up the ruins of some ancient structures, long since collapsed. An abandoned barn on the site, built over what may once have been the homes of herders between pastures, fell into decay around the 1700s or 1800s CE, and fields where stone houses and graves once stood were being plowed and farmed during the Soviet era in 1972.Antiquity, 2017. DOI: 10.15184/aqy.2024.197 (About DOIs).Kiona N. SmithScience correspondentKiona N. SmithScience correspondent Kiona is a freelance science journalist and resident archaeology nerd at Ars Technica. 1 Comments

John Edwards, Technology Journalist & AuthorJanuary 9, 20256 Min ReadPanther Media GmbH via Alamy Stock PhotoDisaster recovery technologies are designed to prevent or minimize the data loss and business disruption resulting from unexpected catastrophic events. This includes everything from hardware failures and local power outages to cyberattacks, natural disasters, civil emergencies, and criminal or military assaults.As AI continues to transform and enhance a seemingly endless array of tasks and functions, it should come as no surprise that the technology has caught the attention of disaster recovery professionals.Preparation and ResponseJoseph Ours, AI strategy director at Centric Consulting, says AI can assist disaster recovery in two essential areas: preparation and response. "In many respects, speeding disaster recovery means planning and preparing," he observes in an email interview. Ours notes that a growing number of government agencies and insurance companies are already routinely performing these tasks with AI assistance. "They use predictive and classification models to analyze historical data and environmental factors to determine potential risk."AI-enabled resiliency planning provides speed and precision that traditional methods lack, says Stephen DeAngelis, president of Enterra Solutions, an AI-enabled transformation and intelligent enterprise planning platform provider. "AI's ability to process large volumes of data quickly allows it to detect anomalies and potential risks earlier," he explains in an online interview. Unlike conventional disaster recovery plans, AI-powered solutions are adaptive, updating in real-time as conditions change. "This means companies can pivot their strategies almost immediately, reducing the time needed to return to normal operations and ensuring minimal disruption to the supply chain."Related:Automatic DetectionIn businesses, AI-enhanced disaster recovery automatically detects anomalies, such as ransomware-corrupted data, allowing technicians to skip over unusable files and focus on clean, viable backups, says Stefan Voss, a vice president at data protection and security firm N-able. "This eliminates the time-intensive, manual review process thats standard in conventional recovery methods."AI can also improve boot detection accuracy, ensuring that machines will bounce back successfully after recovery, Voss says in an email interview. "Well-trained AI models can significantly reduce false positives or negatives, enhancing technician confidence in the reliability and efficiency of the restored systems," he explains. "With AI-driven accuracy, organizations can recover systems faster, with fewer errors, and minimize downtime."Related:AI solutions rely on access to high-quality data to generate accurate predictions. "When data is siloed or incomplete, models are likely to produce less reliable results," DeAngelis warns. To ensure success, he advises businesses to establish robust data management practices before implementing AI solutions. "Today, we're seeing innovators develop sophisticated techniques, such as advanced data modeling, to bridge critical data gaps and enhance AI accuracy."Getting StartedAn important first step toward using AI in disaster recovery is conducting a comprehensive assessment of current supply chain vulnerabilities. "Identify critical points of failure and gather historical data on past disruptions," DeAngelis suggests. Next, collaborate with an AI partner to build predictive models that simulate various disaster scenarios, such as geopolitical risks or extreme weather events. Focus on implementing AI tools that integrate seamlessly with existing systems, allowing for smooth data flows and real-time updates. "A phased approach is ideal, beginning with pilot projects and scaling up as the organization gains familiarity with the technology."Related:Voss says the next step should be identifying any existing challenges in the disaster recovery process. "For example, if your main goal is increasing recovery testing accuracy, look for AI tools designed to improve boot detection and guarantee reliable system restoration," he suggests. "On the other hand, if the goal is precisely detecting backup anomalies, focus on AI solutions that specialize in identifying compromised or corrupted data quickly and accurately."After clearly defining the issue at hand, seek out the AI solution that will meet your needs, Voss advises. "Always start with your pain points and let AI provide the answer, not the other way around."ChallengesAI disaster recovery can offer significant advantages, yet it also comes with several serious drawbacks. High development and integration costs can be a barrier, especially for small businesses, Voss says. "The skills shortage in AI expertise makes it difficult for organizations to develop or maintain AI-driven systems."Remember, too, that even with well-trained models, AI is far from infallible. False positives or negatives can occur, potentially complicating recovery efforts, Voss warns. "Additionally, an over-reliance on AI can reduce human oversight, making it imperative to strike a balance between automation and manual processes."Perhaps the biggest drawback is that some disasters arrive as unpredictable black swan-type events. "In this case, AI is neither a benefit nor contributor to the failure to respond because, by their very nature, humans would struggle to respond adequately as well," Ours says.A Competitive EdgeA proactive investment in AI not only mitigates risk but can turn challenges into competitive advantages, DeAngelis says. He notes that by being prepared to adapt quickly when disruptions occur, enterprises can maintain continuity and even capture market share from less-prepared competitors. "As we've seen from recent events, such as the US port strike, hurricane-related supply chain impacts, and the ongoing pressures of inflation, businesses that leverage AI to build resilience are better positioned to thrive in uncertain environments."About the AuthorJohn EdwardsTechnology Journalist & AuthorJohn Edwards is a veteran business technology journalist. His work has appeared in The New York Times, The Washington Post, and numerous business and technology publications, including Computerworld, CFO Magazine, IBM Data Management Magazine, RFID Journal, and Electronic Design. He has also written columns for The Economist's Business Intelligence Unit and PricewaterhouseCoopers' Communications Direct. John has authored several books on business technology topics. His work began appearing online as early as 1983. Throughout the 1980s and 90s, he wrote daily news and feature articles for both the CompuServe and Prodigy online services. His "Behind the Screens" commentaries made him the world's first known professional blogger.See more from John EdwardsNever Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.SIGN-UPYou May Also LikeWebinarsMore WebinarsReportsMore Reports

Frank Kim, SANS Institute Fellow January 9, 20255 Min ReadEmre Akkoyun via Alamy StockAdopting zero-trust security architectures is increasingly becoming a corporate imperative, with zero trust serving as the recommended approach for building resilience against the evolving nature of enterprise threats. This shift represents more than just implementing the latest and greatest best-of-breed tools. Its a foundational shift away from perimeter-based security controls and external network defenses that were not designed for todays threat landscape.More than 80% of all data breaches today are attributed to human error or negligence, making human risk a pressing security concern amid the rise of hybrid work environments. A zero-trust architecture limits the damage that a compromised user can cause by segmenting the organizations security environment into smaller, isolated zones that restrict the ability to access sensitive data across the entire ecosystem. Unfortunately, the path to effective implementation has proven challenging. Forrester research found that more than 63% of enterprises are struggling to implement zero-trust frameworks, and Gartner predicts that by 2026 only 10% of large enterprises will have a mature and measurable zero-trust program in place.This heightens the role of the transformational CISO to the forefront. CISO success today requires more than being a pure technologist from the SOC room. They need to serve as transformational leaders who are capable of navigating shifting organizational priorities to foster collective buy-in amongst executive leaders, establish effective processes with business line stakeholders, and develop versatile security teams. Cultivating this company-wide alignment is critical to alleviating the roadblocks that hinder zero-trust adoption today.Related:Articulating Zero Trusts ValueNearly 50% of IT professionals describe collaboration between security risk management and business risk management as poor or nonexistent, according to NIST research. As CISOs, its our job to bridge this divide by framing zero trust as an enabler of business agility, operational efficiency, and competitive advantage rather than focusing on technical specifications. Leveraging scenario-based planning and risk quantification techniques can effectively articulate the value of zero trust in terms that resonate with various stakeholders -- correlating the ramifications of cyber incidents to high-value outcomes that impact their department. Marketing leaders, for example, might better appreciate zero trust when they understand how it prevents customer data breaches that result in brand reputational damage.Related:CISOs should establish regular touchpoints with business unit leaders to understand their workflows, pain points, and growth initiatives. This collaborative approach helps identify opportunities where zero trust can enhance business processes rather than hinder them. By securing visible support from the C-suite, CISOs can overcome initial resistance and ensure the necessary resources are allocated for successful implementation. It also helps strengthen organizational buy-in across all employees, giving the company a platform to address concerns, share implementation progress, and maintain alignment with business objectives.Minimizing Organizational FrictionSuccessful zero-trust adoption requires a carefully orchestrated change management strategy. Rather than pursuing lower-risk areas, organizations often achieve better results by starting with mid-risk priorities and moving methodically toward more complex challenges. This approach prevents implementation paralysis and drives meaningful security advancement.Clear communication at every stage is essential. Regular updates, user awareness training, and open feedback channels help maintain transparency and address concerns proactively. When employees realize that zero trust can streamline their access to resources while maintaining security, resistance typically diminishes. The key lies in balancing security requirements with user experience. Modern implementations should leverage automation and contextual access controls to make security seamless. Implementing single sign-on solutions alongside zero-trust principles can enhance both security and convenience, making the transition more palatable for end users.Related:In addition, developing a comprehensive change impact assessment helps identify potential friction points before they emerge. This involves mapping current workflows, understanding dependencies, and creating mitigation strategies. Regular user satisfaction surveys and feedback sessions enable continuous refinement of the implementation approach, ensuring that security measures align with operational needs while maintaining robust protection.Positioning Practitioners for SuccessThe technical complexity of zero-trust architectures demands a targeted focus on skill development amongst security practitioners. With practitioners often wearing multiple hats across architecture, implementation, operations, and monitoring, they must be all-around defenders who are capable of seamlessly transitioning between functional roles. This requires a strong foundational knowledge spanning both on-premises and cloud security domains. Security teams must understand the organization's end-to-end security environment, from network tools to cloud applications, endpoints, and data storage systems.Investment in targeted learning is crucial here. Prioritize formal trainings and upskilling programs that build team-wide competencies and implement cross-training initiatives that facilitate knowledge sharing to reduce key person dependencies and develop operational resilience. Establishing a dedicated zero-trust center of excellence can accelerate this skill development by providing guidance and support to other security team members while maintaining documentation and best practices.The path to zero trust is a continuous journey of organizational transformation. While technical implementation remains crucial, the transformational CISO's ability to bridge cultural gaps, foster organizational alignment, and develop comprehensive team capabilities will determine the success of zero-trust initiatives. As cyber threats continue to evolve and regulatory pressures mount, organizations that successfully execute this cultural and technical transformation will be better positioned to protect their critical assets and maintain business continuity in an increasingly complex threat landscape.About the AuthorFrank KimSANS Institute Fellow Frank Kim is a SANS Fellow where he leads the Cloud Security and Cybersecurity Leadership curricula to help shape and develop the next generation of security leaders. Previously, he served as the organizations CISO where he led the information risk function for the most trusted source of cybersecurity training and certification in the world.See more from Frank KimNever Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.SIGN-UPYou May Also LikeWebinarsMore WebinarsReportsMore Reports

AI tools will enable significant productivity and efficiency benefits for organizations in the coming year, but they also will exacerbate privacy, governance, and security risks.

Quantum computers could use heat to eliminate errorsChalmers University of Technology, Lovisa HkanssonA tiny cooling device can automatically reset malfunctioning components of a quantum computer. Its performance suggests that manipulating heat could also enable other autonomous quantum devices.Quantum computers arent yet fully practical because they make too many errors. In fact, if qubits key components of this type of computer accidentally heat up and become too energetic, they can end up in an erroneous state before the calculation even begins. One way to reset the qubits to their correct states is to cool them down. AdvertisementSimone Gasparinetti at Chalmers University of Technology in Sweden and his colleagues have delegated this task to an autonomous quantum refrigerator for the first time.The researchers built two qubits and one qutrit, which can store more complex information than a qubit, from tiny superconducting circuits. The qutrit and one of the qubits formed a fridge for the second target qubit, which could eventually be used for computation.The researchers carefully engineered the interactions between the three components to ensure that when the target qubit had too much energy, which caused errors, heat automatically flowed out of it and into the two other elements. This lowered the target qubits temperature and reset it. Because this process was autonomous, the qubit-and-qutrit fridge could correct errors without any outside control. Receive a weekly dose of discovery in your inbox.Sign up to newsletterAamir Ali, also at Chalmers University of Technology, says this approach to resetting the qubit required less new hardware than more conventional methods and yielded better results. Without any significant quantum computer redesign or introduction of new wires, the qubits starting state was correct 99.97 per cent of the time. In contrast, other reset methods typically only manage 99.8 per cent, he says.This is a powerful example of how thermodynamic machines which deal with heat, energy and temperature can be useful in the quantum realm, says Nicole Yunger Halpern at the National Institute of Standards and Technology in Maryland, who worked on the project.Conventional thermodynamic machines like the heat engine sparked a whole industrial revolution, but so far quantum thermodynamics hasnt been very practical. Im interested in making quantum thermodynamics useful. This arguably useful autonomous quantum refrigerator is our first example, says Yunger Halpern.Its nice to see this machine implemented and useful. The fact that it is autonomous, so it does not require any external control, should make it efficient and versatile, says Nicolas Brunner at the University of Geneva in Switzerland.Micha Horodecki at the University of Gdask in Poland says that one of the most urgent problems for quantum computers built with superconducting circuits is making sure the machines dont heat up and subsequently make errors. The new experiment opens a path for many similar projects that have been proposed but never tested, such as using qubits to build autonomous quantum engines, he says.The researchers are already looking into whether they could build on their experiment. For example, they might create an autonomous quantum clock or design a quantum computer with other functions automatically driven by temperature differences.Journal referenceNature Physics DOI: http://dx.doi.org/10.1038/s41567-024-02708-5Topics:

Comment and TechnologyWe need to stop ignoring young people's firsthand experience with artificial intelligence. They are already at the sharp end of its development, says Mhairi Aitken 8 January 2025 Adri VoltNext month, world leaders will gather in Paris for the AI Action Summit, the latest in a series of high-profile international events on artificial intelligence. Previous ones have brought together heads of state, senior policy-makers and CEOs of tech companies to discuss how to tackle the risks of advanced AI technologies. But there is one group that has so far been entirely missing from these processes, and it happens to be the one that will be most affected by advances in AI: children.Children are consistently the most underrepresented group in decision-making processes relating to AIs design, development and deployment,

This article is from The Spark, MIT Technology Reviews weekly climate newsletter. To receive it in your inbox every Wednesday, sign up here. I love the fresh start that comes with a new year. And one thing adding a boost to my January is our newest list of 10 Breakthrough Technologies. In case you havent browsed this years list or a previous version, it features tech thats either breaking into prominence or changing society. We typically recognize a range of items running from early-stage research to consumer technologies that folks are getting their hands on now. As I was looking over the finished list this week, I was struck by something: While there are some entries from other fields that are three or even five years away, all the climate items are either newly commercially available or just about to be. Its certainly apt, because this year in particular seems to be bringing a new urgency to the fight against climate change. Were facing global political shifts and entering the second half of the decade. Its time for these climate technologies to grow up and get out there. Green steel Steel is a crucial material for buildings and vehicles, and making it accounts for around 8% of global greenhouse-gas emissions. New manufacturing methods could be a huge part of cleaning up heavy industry, and theyre just on the cusp of breaking into the commercial market. One company, called Stegra, is close to starting up the worlds first commercial green steel plant, which will make the metal using hydrogen from renewable sources. (You might know this company by its former name, H2 Green Steel, as we included it on our 2023 list of Climate Tech Companies to Watch.) When I first started following Stegra a few years ago, its plans for a massive green steel plant felt incredibly far away. Now the company says its on track to produce steel at the factory by next year. The biggest challenge in this space is money. Building new steel plants is expensiveStegra has raised almost $7 billion. And the companys product will be more expensive than conventional material, so itll need to find customers willing to pay up (so far, it has). There are other efforts to clean up steel that will all face similar challenges around money, including another play in Sweden called Hybrit and startups like Boston Metal and Electra, which use different processes. Read more about green steel, and the potential obstacles it faces as we enter a new phase of commercialization, in this short blurb and in this longer feature about Stegra. Cow burp remedies Humans love burgers and steaks and milk and cheese, so we raise a whole bunch of cows. The problem is, these animals are among a group with a funky digestion process that produces a whole lot of methane (a powerful greenhouse gas). A growing number of companies are trying to develop remedies that help cut down on their methane emissions. This is one of my favorite items on the list this year (and definitely my favorite illustrationat the very least, check out this blurb to enjoy the art). Theres already a commercially available option right now: a feed additive called Bovaer from DSM-Firmenich that the company says can cut methane emissions by 30% in dairy cattle, and more in beef cattle. Startups are right behind with their own products, some of which could prove even better. A key challenge all these companies face moving forward is acceptance: from regulatory agencies, farmers, and consumers. Some companies still need to go through lengthy and often expensive tests to show that their products are safe and effective. Theyll also need to persuade farmers to get on board. Some might also face misinformation thats causing some consumers to protest these new additives. Cleaner jet fuel While planes crisscrossing the world are largely powered by fossil fuels, some alternatives are starting to make their appearance in aircraft. New fuels, today mostly made from waste products like used cooking oil, can cut down emissions from air travel. In 2024, they made up about 0.5% of the fuel supply. But new policies could help these fuels break into new prominence, and new options are helping to widen their supply. The key challenge here is scale. Global demand for jet fuel was about 100 billion gallons last year, so well need a whole lot of volume from new producers to make a dent in aviations emissions. To illustrate the scope, take LanzaJets new plant, opened in 2024. Its the first commercial-scale facility that can make jet fuel with ethanol, and it has a capacity of about 9 million gallons annually. So we would need about 10,000 of those plants to meet global demanda somewhat intimidating prospect. Read more in my write-up here. From cow burps to jet fuel to green steel, theres a huge range of tech thats entering a new stage of deployment and will need to face new challenges in the next few years. Well be watching it allthanks for coming along. Now read the rest of The Spark Related reading Check out our full list of 2025s Breakthrough Technologies here. Theres also a poll where you can vote for what you think the 11th item should be. Im not trying to influence anyones vote, but I think methane-detecting satellites are pretty interestingjust saying This package is part of our January/February print issue, which also includes stories on: This system thats tracking early warning signs of infection in wheat crops How wind could be a low-tech solution to help clean up shipping Efforts to use human waste in agriculture JUSTIN SULLIVAN/GETTY Another thing EVs are (mostly) set for solid growth in 2025, as my colleague James Temple covers in his newest story. Check it out for more about whats next for electric vehicles, including what we might expect from a new administration in the US and how China is blowing everyone else out of the water. Keeping up with climate Winter used to be the one time of year that California didnt have to worry about wildfires. A rapidly spreading fire in the southern part of the state is showing thats not the case anymore. (Bloomberg) Teslas annual sales decline for the first time in over a decade. Deliveries were lower than expected for the final quarter of the year. (Associated Press) Meanwhile, in China, EVs are set to overtake traditional cars in sales years ahead of schedule. Forecasts suggest that EVs could account for 50% of car sales this year. (Financial Times) KoBold metals raised $537 million in funding to use AI to mine copper. The funding pushes the startups valuation to $2.96 billion. (TechCrunch) Read this profile of the company from 2021 for more. (MIT Technology Review)We finally have the final rules for a tax credit designed to boost hydrogen in the US. The details matter here. (Heatmap) China just approved the worlds most expensive infrastructure project. The hydroelectric dam could produce enough power for 300 million people, triple the capacity of the current biggest dam. (Economist) In 1979, President Jimmy Carter installed 32 solar panels on the White Houses roof. Although they came down just a few years later, the panels lived multiple lives afterward. I really enjoyed reading about this small piece of Carters legacy in the wake of his passing. (New York Times) An open pit mine in California is the only one in the US mining and extracting rare earth metals including neodymium and praseodymium. This is a fascinating look at the site. (IEEE Spectrum) I wrote about efforts to recycle rare earth metals, and what it means for the long-term future of metal supply, in a feature story last year. (MIT Technology Review)

The first Democrat in New York history with a computer science background wants to revive some of the ideas behind the failed California AI safety bill, SB 1047, with a new version in his state that would regulate the most advanced AI models. Its called the RAISE Act, an acronym for Responsible AI Safety and Education. Assembly member Alex Bores hopes his bill, currently an unpublished draftsubject to changethat MIT Technology Review has seen, will address many of the concerns that blocked SB 1047 from passing into law. SB 1047 was, at first, thought to be a fairly modest bill that would pass without much fanfare. In fact, it flew through the California statehouse with huge margins and received significant public support. However, before it even landed on Governor Gavin Newsoms desk for signature in September, it sparked an intense national fight. Google, Meta, and OpenAI came out against the bill, alongside top congressional Democrats like Nancy Pelosi and Zoe Lofgren. Even Hollywood celebrities got involved, with Jane Fonda and Mark Hamill expressing support for the bill. Ultimately, Newsom vetoed SB 1047, effectively killing regulation of so-called frontier AI models not just in California but, with the lack of laws on the national level, anywhere in the US, where the most powerful systems are developed. Now Bores hopes to revive the battle. The main provisions in the RAISE Act include requiring AI companies to develop safety plans for the development and deployment of their models. The bill also provides protections for whistleblowers at AI companies. It forbids retaliation against an employee who shares information about an AI model in the belief that it may cause critical harm; such whistleblowers can report the information to the New York attorney general. One way the bill defines critical harm is the use of an AI model to create a chemical, biological, radiological, or nuclear weapon that results in the death or serious injury of 100 or more people. Alternatively, a critical harm could be a use of the AI model that results in 100 or more deaths or at least $1 billion in damages in an act with limited human oversight that if committed by a human would constitute a crime requiring intent, recklessness, or gross negligence. The safety plans would ensure that a company has cybersecurity protections in place to prevent unauthorized access to a model. The plan would also require testing of models to assess risks before and after training, as well as detailed descriptions of procedures to assess the risks associated with post-training modifications. For example, some current AI systems have safeguards that can be easily and cheaply removed by a malicious actor. A safety plan would have to address how the company plans to mitigate these actions. The safety plans would then be audited by a third party, like a nonprofit with technical expertise that currently tests AI models. And if violations are found, the bill empowers the attorney general of New York to issue fines and, if necessary, go to the courts to determine whether to halt unsafe development. A different flavour of bill The safety plans and external audits were elements of SB 1047, but Bores aims to differentiate his bill from the California one. We focused a lot on what the feedback was for 1047, he says. Parts of the criticism were in good faith and could make improvements. And so we've made a lot of changes. The RAISE Act diverges from SB 1047 in a few ways. For one, SB 1047 would have created the Board of Frontier Models, tasked with approving updates to the definitions and regulations around these AI models, but the proposed act would not create a new government body. The New York bill also doesnt create a public cloud computing cluster, which SB 1047 would have done. The cluster was intended to support projects to develop AI for the public good. The RAISE Act doesnt have SB 1047s requirement that companies be able to halt all operations of their model, a capability sometimes referred to as a kill switch. Some critics alleged that the shutdown provision of SB 1047 would harm open-source models, since developers cant shut down a model someone else may now possess (even though SB 1047 had an exemption for open-source models). The RAISE Act avoids the fight entirely. SB 1047 referred to an advanced persistent threat associated with bad actors trying to steal information during model training. The RAISE Act does away with that definition, sticking to addressing critical harms from covered models. Focusing on the wrong issues? Bores bill is very specific with its definitions in an effort to clearly delineate what this bill is and isnt about. The RAISE Act doesnt address some of the current risks from AI models, like bias, discrimination, and job displacement. Like SB 1047, it is very focused on catastrophic risks from frontier AI models. Some in the AI community believe this focus is misguided. Were broadly supportive of any efforts to hold large models accountable, says Kate Brennan, associate director of the AI Now Institute, which conducts AI policy research. But defining critical harms only in terms of the most catastrophic harms from the most advanced models overlooks the material risks that AI poses, whether its workers subject to surveillance mechanisms, prone to workplace injuries because of algorithmically managed speed rates, climate impacts of large-scale AI systems, data centers exerting massive pressure on local power grids, or data center construction sidestepping key environmental protections," she says. Bores has worked on other bills addressing current harms posed by AI systems, like discrimination and lack of transparency. That said, Bores is clear that this new bill is aimed at mitigating catastrophic risks from more advanced models. Were not talking about any model that exists right now, he says. We are talking about truly frontier models, those on the edge of what we can build and what we understand, and there is risk in that. The bill would cover only models that pass a certain threshold for how many computations their training required, typically measured in FLOPs (floating-point operations). In the bill, a covered model is one that requires more than 1026 FLOPs in its training and costs over $100 million. For reference, GPT-4 is estimated to have required 1025 FLOPs. This approach may draw scrutiny from industry forces. While we cant comment specifically on legislation that isnt public yet, we believe effective regulation should focus on specific applications rather than broad model categories, says a spokesperson at Hugging Face, a company that opposed SB 1047. Early days The bill is in its nascent stages, so its subject to many edits in the future, and no opposition has yet formed. There may already be lessons to be learned from the battle over SB 1047, however. Theres significant disagreement in the space, but I think debate around future legislation would benefit from more clarity around the severity, the likelihood, and the imminence of harms, says Scott Kohler, a scholar at the Carnegie Endowment for International Peace, who tracked the development of SB 1047. When asked about the idea of mandated safety plans for AI companies, assembly member Edward Ra, a Republican who hasn't yet seen a draft of the new bill yet, said: I dont have any general problem with the idea of doing that. We expect businesses to be good corporate citizens, but sometimes you do have to put some of that into writing. Ra and Bores co chair the New York Future Caucus, which aims to bring together lawmakers 45 and under to tackle pressing issues that affect future generations. Scott Wiener, a California state senator who sponsored SB 1047, is happy to see that his initial bill, even though it failed, is inspiring further legislation and discourse. The bill triggered a conversation about whether we should just trust the AI labs to make good decisions, which some will, but we know from past experience, some wont make good decisions, and thats why a level of basic regulation for incredibly powerful technology is important, he says. He has his own plans to reignite the fight: Were not done in California. There will be continued work in California, including for next year. Im optimistic that California is gonna be able to get some good things done. And some believe the RAISE Act will highlight a notable contradiction: Many of the industrys players insist that they want regulation, but when any regulation is proposed, they fight against it. SB 1047 became a referendum on whether AI should be regulated at all, says Brennan. There are a lot of things we saw with 1047 that we can expect to see replay in New York if this bill is introduced. We should be prepared to see a massive lobbying reaction that industry is going to bring to even the lightest-touch regulation. Wiener and Bores both wish to see regulation at a national level, but in the absence of such legislation, theyve taken the battle upon themselves. At first it may seem odd for states to take up such important reforms, but California houses the headquarters of the top AI companies, and New York, which has the third-largest state economy in the US, is home to offices for OpenAI and other AI companies. The two states may be well positioned to lead the conversation around regulation. There is uncertainty at the direction of federal policy with the transition upcoming and around the role of Congress, says Kohler. It is likely that states will continue to step up in this area. Wieners advice for New York legislators entering the arena of AI regulation? Buckle up and get ready.

Hundreds of incarcerated firefighters are helping to battle the wildfires in the Los Angeles area.Incarcerated firefighters earn $26.90 to $34 for each 24-hour shift.It's far below California's minimum wage of $16.50 an hour.Hundreds of incarcerated firefighters have been deployed to help battle the fires raging across the Los Angeles area, with some working 24-hour shifts for as little as $26.90, or just over $1 an hour.The California Department of Corrections and Rehabilitation told The Guardian that it had deployed 395 incarcerated firefighters to help battle the blazes.The firefighters, who have "minimum custody" status, have been embedded with the California Department of Forestry and Fire Protection, which has deployed thousands of firefighting to the area.The CDCR, jointly with Cal Fire and the Los Angeles County Fire Department, runs 35 "fire camps" across 25 California counties. Participation is voluntary, with inmates using hand tools to aid in fire suppression and other emergency responses.The camps are considered minimum-security facilities.California, which is grappling with longer and more destructive fire seasons amid the climate crisis, has long relied on incarcerated people for its response. In fact, the CDCR crews have, at times, accounted for as much as 30% of the state's wildfire force.According to the CDCR, fire crew members earn between $5.80 and $10.24 a day, depending on their skill level, with an additional $1 per hour for active emergency assignments.Crews can work 24-hour shifts during emergencies, followed by 24 hours of rest, it said, with the lowest-skilled firefighters earning $26.90 and the highest-skilled maxing out at about $34.For context, California's minimum wage is $16.50 per hour, with some areas, such as West Hollywood, offering higher minimum wages.In California, inmates are not guaranteed the state's minimum wage, and some earn as little as 16 cents per hour.The $10.24 basic rate for the highest-skilled incarcerated firefighters ranks among the best daily rate for incarcerated people in the state.In addition to pay, fire crews helping out during emergencies like this receive "time credits" on a two-for-one basis, meaning that for each day they serve on the crew, they receive two additional days off their sentence.Participation can also lead to criminal record expungement and the ability to seek professional emergency response certifications.The CDCR did not immediately respond to Business Insider's request for comment.

American Airlines flight attendants were sent a union memo about leaving passengers on empty planes.The AFPA said cabin crew members should check lavatories and under seats for sleeping passengers.A union spokesperson said it "routinely" issued such reminders to members.The union for American Airlines flight attendants has issued a memo reminding cabin crew not to leave sleeping passengers on board after flights land.Aviation news site Paddle Your Own Kanoo reported that the Association of Professional Flight Attendants (APFA) recently warned members there had been an increase in such incidents."This communication serves as a reminder for flight attendants to complete required post-flight security checks to ensure all passengers have deplaned upon arrival," a spokesperson for APFA, which represents more than 28,000 American Airlines cabin crew members, told Business Insider.They added that the union "routinely releases communications reminding flight attendants of the importance of conducting these checks."Leaving a passenger on an empty plane is not only a security concern but can also violate safety rules that mandate a minimum number of flight attendants per passenger.APFA's recent memo reminded cabin crew to "check lavatories, and make sure no passengers are sleeping in or under seats," Paddle Your Own Kanoo reported."These procedures are required to ensure that no passenger is ever left onboard," the memo continued. "This is a critical final check to confirm that no one has been overlooked, and this ensures that you and your crew aren't at risk of violating a [federal aviation regulation]."The lead flight attendant is also supposed to perform a walk-through of the cabin as a final check, the memo reportedly said."Leaving passengers on the aircraft unattended is a significant safety and security concern and we appreciate everyone's shared efforts in ensuring this doesn't happen."In 2019, an Air Canada passenger woke up alone on a cold and dark plane.Tiffani O'Brien said she fell asleep on the 90-minute flight from Quebec City to Toronto and woke up hours later after the plane had been parked.