DatacenterPixabayAt Microsoft Ignite 2024, the company unveiled a series of advancements that signify a strategic shift towards autonomous AI agents, aiming to enhance operational efficiency and productivity across various sectors. Central to this initiative is the integration of Copilot with agent-based systems, reflecting Microsofts move in transforming assistive AI tools into autonomous agents capable of performing complex tasks with minimal human intervention.1. Azure AI Foundry: The Unified Development PlatformMicrosoft has rebranded Azure AI Studio as Azure AI Foundry. It is a unified platform designed to streamline the development, customization and management of AI applications. It integrates various Azure AI services and tools, providing developers with a comprehensive environment to build and deploy AI solutions efficiently. The platform includes a new software development kit that facilitates integration with familiar development environments like GitHub and Visual Studio, promoting seamless collaboration and innovation.Azure AI Foundry uses a hub-and-project architecture, where the hub serves as the top-level resource managing security configurations, compute resources and service connections, while projects are child resources that provide isolated development environments with access to tools, reusable components and specific project-scoped connections. The platform emphasizes centralized governance, enabling teams to efficiently manage security, connectivity and computing resources across multiple projects while maintaining granular access control through Azure role-based access control and attribute-based access control.The platform enables developers to manage the end-to-end lifecycle of generative AI applications through model selection, fine-tuning, deployment, retrieval-augmented generation, guardrails and governance.2. Azure AI Agent Service: Autonomous AI FrameworkMicrosofts Azure AI Agent Service is a capability of Azure AI Foundry for developers to create, deploy and scale intelligent AI agents that can automate complex business processes. The service enables developers to build secure, stateful autonomous agents by integrating models and technologies from Microsoft, OpenAI and partners like Meta, Mistral and Cohere. These agents can leverage knowledge from diverse sources, including Bing, SharePoint, Fabric, Azure AI Search, Azure Blob and licensed data repositories, providing unprecedented flexibility in agent development.Read More: Azure Local Brings The Power Of Cloud To On-Premises And EdgeThe Azure AI Agent Service introduces managed capabilities that simplify AI agent creation, allowing organizations to develop purpose-built solutions that can handle intricate workflows with minimal manual intervention. Developers can use a code-first approach to customize AI solutions, enabling agents to work across multiple data platforms and integrate seamlessly with existing systems. The service supports autonomous agents that can plan, learn from processes, adapt to new conditions and make decisions independently, effectively transforming how businesses approach task automation and operational efficiency.Azure AI Agent Service integrates seamlessly with Logic Apps, Power Apps and Azure Functions, enabling developers to create sophisticated AI-driven applications. By leveraging Azure Functions, developers can implement custom logic and actions within AI agents, facilitating complex workflows and real-time data processing. This integration allows AI agents to perform tasks such as sending emails, scheduling meetings and automating report creation. Azure Logic Apps provide a powerful mechanism for integrating with the Azure AI Agent SDK through function calling capabilities. The integration enables developers to create intelligent, automated workflows that can be dynamically invoked by AI agents. Additionally, Power Apps provides a low-code platform for building user interfaces that interact with these AI agents, allowing users to engage with AI-driven functionalities through intuitive applications.This synergy between Azure AI Agent Service, Logic Apps, Power Apps and Azure Functions empowers organizations to develop intelligent, automated solutions tailored to their specific business needs. For orchestrating multiple agents, Microsoft has plans to integrate Autogen, a powerful open-source framework for agentic workflows.3. Copilot Studio + Azure AI Foundry: Bridging Assistant and Agent CapabilitiesMicrosoft Copilot and Azure AI Agents represent two distinct approaches within Microsofts AI ecosystem, each serving unique functions to enhance user productivity. Microsoft 365 Copilot acts as an AI-powered assistant embedded within applications like Microsoft 365, providing real-time assistance, generating content and offering contextual suggestions to users. In contrast, agents are autonomous AI entities designed to perform tasks independently, automating complex workflows and processes without continuous user input.Microsoft Copilot Studio targets knowledge workers to create agents in natural language, while the new AI Foundry Agent SDK is meant for developers and builders creating sophisticated and autonomous agentic workflows.At Ignite 2024, Microsoft showcased how it plans to bridge the gap between the two. The Copilot Studio now offers autonomous agentic capabilities, allowing makers to build agents that can take actions independently, such as responding to emails or recording uploaded files without constant human prompting. The new Agent SDK empowers developers to create multi-channel agents leveraging Azure AI, Semantic Kernel and Copilot Studio services, deployable across platforms like Teams, Copilot, web and third-party messaging systems.The integration between Copilot Studio and AI Foundry Agents introduces features like an agent library with templates for common scenarios, including leave management, sales order processing and deal acceleration. Developers can now build full-stack, trusted agents with access to the Copilot Trust Layer, enabling seamless integration between low-code and pro-code solutions. Additional capabilities include image upload for agent analysis, voice-enabled agent creation and advanced knowledge tuning. Documents indexed in Azure AI Foundry can be used in Copilot Studio as knowledge sources for agents. The integration also provides IT professionals with a Copilot Control System to securely manage agent functionalities, ensuring enterprises can customize and deploy AI agents that align precisely with their unique business workflows and compliance requirements.4. Azure AI Reports: Enhanced Governance FrameworkAt Microsoft Ignite 2024, Azure AI Reports was announced as a critical tool for enterprises seeking comprehensive insights and governance for their AI initiatives. The platform provides detailed documentation and evaluation mechanisms for AI models, enabling organizations to track model performance, assess potential risks and generate transparent model cards that capture key characteristics and limitations. These reports are designed to support responsible AI development by offering granular visibility into model behaviors, potential biases and performance metrics across different scenarios.Azure AI Reports are integrated into the Azure AI Foundry portal, providing a centralized location for managing AI projects and resources. The enhanced user interface features streamlined navigation, making it easier to discover AI capabilities and manage applications efficiently. Additionally, the portal includes a new management center that allows users to govern projects, resources, deployments and quotas, further supporting the effective oversight of AI initiatives.The Azure AI Reports feature introduces advanced capabilities for enterprises to maintain compliance and ethical standards in AI deployment. By generating automated documentation that covers model training data, performance benchmarks and potential use case limitations, organizations can now create a structured approach to AI governance. The platform integrates seamlessly with existing Azure AI services, allowing developers and IT professionals to access comprehensive insights directly through familiar tools like GitHub and Visual Studio, thereby simplifying the process of maintaining transparency and accountability in AI model development.5. Serverless GPU Computing: Infrastructure Evolution for AIAzure Container Apps is a fully managed serverless container service that enables developers to build and deploy modern, cloud-native applications and microservices at scale.At Microsoft Ignite 2024, the platform introduced serverless GPU support, a groundbreaking feature that allows developers to access NVIDIA A100 and T4 GPUs without managing complex infrastructure. This capability provides a flexible, pay-per-second compute option that scales automatically, eliminating the traditional overhead of GPU resource management.The serverless GPU support offers critical advantages for AI and machine learning developers. By providing scale-to-zero capabilities, developers can run GPU-intensive workloads like model training, inference and video rendering without maintaining dedicated hardware. The feature supports full data governance, ensuring that data never leaves the container boundary, which is crucial for enterprises with strict security requirements. Developers can choose between NVIDIA A100 and T4 GPU types, offering flexibility for different computational needs while benefiting from per-second billing and automatic scaling.GPU support in Azure Container Apps bridges the gap between serverless APIs and traditional managed compute, making high-performance computing resources more accessible. Developers can now focus on core AI code rather than infrastructure management, with the platform handling complex GPU provisioning and scaling. Currently available in West US 3 and Australia East regions, this feature is particularly transformative for AI development teams seeking a streamlined, secure and scalable approach to GPU-accelerated computing.SummaryThese announcements reflect Microsoft's commitment to enterprise AI deployment at scale. The shift to autonomous agents, combined with consumption-based infrastructure and enhanced governance tools, enables organizations to accelerate AI adoption while maintaining control over costs and risks.Enterprise leaders should evaluate their AI strategy in light of these developments, particularly focusing on opportunities for workflow automation and the transition from fixed to variable AI computing costs.

Storage requirements for AAA games continue to grow each year. While PC gamers have a wide array of storage options and to expand with ease, the situation is a bit more complicated for console gamers on current generation hardware, particularly those using the PlayStation 5, PS5 Pro, and Xbox Series X/S.The Xbox Series X is available with a 1TB SSD, offering approximately 800GB of usable space for games and apps, and a 2TB version with 1.6TB accessible to the user. The Series S, on the other hand, comes in a 1TB model or a 512GB version, the latter providing just 364GB of usable storage. Beyond this, both consoles support running current-gen games exclusively from official expansion cards manufactured by Seagate or WD. However, they can play Xbox One titles from any USB 3.0 drive and allow transferring new games to and from these external drives for storage purposes.The original PlayStation 5 features an 825GB SSD, of which 667GB is available for games and apps. The PS5 slim model, released in 2023, increased storage to 1TB, with approximately 840GB usable. The PS5 Pro, meanwhile, includes a 2TB SSD, with almost 1.9TB accessible to users.In terms of upgradability, the PS5 series offers more flexibility than Xbox consoles, thanks to its internal M.2 slot. However, this comes with notable constraints. Some SSDs are officially endorsed as PS5-compatible, while others are marketed as compatible by their manufacturers. Similar to Xbox consoles, the PS5 can play previous-gen games from a USB drive or use the drive for storing current-gen titles.The PlayStation 5 does not support SATA drives, even those in the M.2 form factor. Moreover, it requires SSDs to support PCIe 4.0 speeds and four PCIe lanes, with a recommended sequential read speed of 5,500MB/s. The PS5 also lacks support for host memory buffer (HMB), meaning SSDs without dedicated DRAM may experience slower game load times and a reduced lifespan.Additionally, the dimensions of the drive's heatsink are strictly regulated: it must be no wider than 25mm (only 3mm more than the drive itself), no taller than 2.45mm beneath the drive, and no taller than 8mm above it. If you have a compatible drive but lack a suitable heatsink, you can purchase a generic compatible heatsink for as little as $7 (or $9 if it's a high-capacity, double-sided drive).Whether you're looking for a drive to expand storage for current-gen Xbox or PlayStation 5 games, or an external drive for legacy titles, this guide has you covered.Best PlayStation 5 SSDsWD Black SN850X | Corsair M600 Pro LPX | Samsung 990 ProIn numbers Price: $135 on AmazonWhen choosing a drive for your PlayStation 5 console, the first factor to consider is capacity. If you need 8TB of storage, the WD Black SN850X is the obvious choice. With denser flash memory than the other versions, it can use the same number of dies as the 4TB model and see no performance loss. An empty 8TB drive will have a big enough writing cache to copy over the entire content of a full PS5 Pro drive at full speed.As of writing, the version without a heatsink is $80 cheaper than the one with a heatsink. You can save money by pairing it with a third-party heatsink that fits double-sided drives, which costs around $9.In case you were wondering, the SN850P is basically the same drive, with a special heatsink design that allows it to display a bigger PlayStation logo, for even more money.For a 4TB drive, the SN850X is still a solid choice, but there are two other options with a single-sided design that simplify cooling and come with reasonably priced heatsinks. For PS5 Pro users, the Corsair MP600 Pro LPX is a better fit due to its superior sustained write speeds. Meanwhile, PS5 users may prefer the Samsung 990 Pro, which offers greater efficiency and slightly faster game load times.If 2TB of storage is sufficient, the single-sided SN850X 2TB is back in the race. Another good option is the SK Hynix P41, though it doesn't include a heatsink. While all of these drives are also available in 1TB versions, they typically offer either lower performance or a much higher cost per terabyte. Back to top Xbox Storage Expansion CardsSeagate Expansion Card or WD Black C50In numbers Price: $129 on AmazonWith the current Xbox lineup, storage expansion couldn't be simpler. Both the WD Black C50 and the Seagate storage expansion card provide identical functionality. Just choose the capacity you want and get the more affordable option available.Both drives are available in 2TB, 1TB, and 500GB versions. If your budget allows, we recommend going for the 2TB card. It not only offers better performance but also provides greater flexibility and convenience compared to buying two separate 1TB drives at different times.While Xbox Storage Expansion Cards are significantly more expensive per GB than M.2 drives, they are far easier to install. This simplicity positions the Xbox as a more user-friendly, family-oriented option, while the PlayStation 5 caters more to power users who value performance. Back to top Best External DriveCrucial X9 Pro Portable SSDIn numbers Price: $80 on AmazonXbox One and PlayStation 4 games were designed to run off a spinning disk, but can still enjoy the faster load times of an SSD. The Crucial X9 Pro stands above the crowd thanks to its dedicated DRAM, especially when the drive houses hundreds of GBs. When connected to the PlayStation 5's USB-C port, it can also transfer games from other drives at 1050MB/s. Unlike its predecessor, the X8 Pro, the X9 Pro rarely drops below this speed.For Xbox users, connecting the Crucial X9 Pro requires a USB-C to USB-A adapter. While transfer speeds are limited to 400MB/s in this setup, game load times remain virtually indistinguishable from those on the PlayStation. The X9 Pro is also built for durability, featuring two-meter drop resistance and an IP55 rating for water and dust protection more than adequate for a drive that is unlikely to move frequently.A good alternative is the Samsung T7 Shield, usually available for a similar price per GB. It doesn't have DRAM, but it does have a USB-A connector. It's just as fast as the Crucial drive for data transfers and it's also rated for three-meter drops, IP65 water and dust resistance.If you're looking to download a huge library of previous-gen AAA titles onto a single drive on the cheap, the Western Digital My Book series (offering several TBs of storage) still offers good value. However, it does have a significant drawback when used with an SSD, as its transfer speeds max out at around 175MB/s, creating a noticeable bottleneck.For those willing to sacrifice even more load-time performance in favor of a compact drive that doesn't require an external power source, the WD Passport series is a reasonable option. However, data transfer speeds are limited to just 130MB/s, which is slower than other options but still sufficient for legacy game storage. Back to top

Artists concept of a near-Earth object. NASA/JPL-Caltech / NASA/JPL-CaltechA so-called mini moon thats been orbiting Earth for the last couple of months is about to disappear off into our solar system again.The near-Earth asteroid, officially called 2024 PT5, was first spotted in early August by the NASA-funded Sutherland, South Africa, telescope of the University of Hawaiis Asteroid Terrestrial-impact Last Alert System (ATLAS). The term mini moon emphasizes the rocks small size its thought to be about 10 meters (33 feet) wide and also its temporary nature as it will never be permanently captured in Earths gravitational field.Recommended VideosThe rocks orbital distance from Earth has been about nine times that of the moons distance from our planet, and it never posed a risk to earthlings.RelatedGiven the similarity between asteroid 2024 PT5s motion and that of our planets, scientists at NASAs Center for Near Earth Object Studies (CNEOS) believe that the object could be a large chunk of rock ejected from the moons surface after an asteroid impact long ago, NASA said.After Monday, November 25, asteroid 2024 PT5 will leave Earths vicinity and continue its journey around the sun. The timing of the rocks exit is influenced by the gravitational interactions between Earth, the moon, and the sun, and it is the gravitational forces exerted by these celestial bodies that is causing the asteroid to be pulled back into its original path around the sun.Professor Carlos de la Fuente Marcos, one of those involved in the discovery of 2024 PT5, told Space.com in September just ahead of the rocks arrival that it belongs to the Arjuna asteroid belt, a secondary asteroid belt made of space rocks that follow orbits very similar to that of Earth at an average distance to the sun of about 93 million miles (150 million kilometers). Commenting on 2024 PT5 becoming a temporary companion of Earth, he added, You may say that if a true satellite is like a customer buying goods inside a store, objects like 2024 PT5 are window shoppers.Besides a close pass in January next year, 2024 PT5 wont be dropping by our neighborhood again until 2055, and after that in 2084.Editors Recommendations

SpaceX could be targeting as many as 25 launches of its Starship rocket for 2025 as it readies the massive vehicle for crew and cargo trips to the moon, Mars, and possibly beyond.The targeted launch cadence for the Starship, which comprises the first-stage Super Heavy booster and the upper-stage Starship spacecraft, appears in a Federal Aviation Administration (FAA) draft environmental assessment for Starship missions from Boca Chica, Texas. The document primarily addresses the environmental considerations and regulatory processes linked to SpaceXs desire to increase the frequency of its Starship test flights from its Starbase facility in Boca Chica.Recommended VideosSpaceXs last stated target, established in 2022 before the rockets first flight in April of the following year, was for as many as five Starship launches per year, but now that the company is getting into a rhythm with its test flights conducting the last two in the span of two months SpaceX wants to ramp up its efforts in order to develop the rocket more rapidly.RelatedThe document states that SpaceX wants to perform up to 25 annual Starship orbital launches, up to 25 annual landings of the Super Heavy booster, and up to 25 annual landings of the Starship spacecraft. SpaceX boss Elon Musk has said that the ultimate goal is to see hundreds of Starship launches taking place each year, though environmentalists believe each Starship launch is damaging the local area.While the launches would originate from SpaceXs Starbase facility, the landings could take place in a number of locations, including at Starbase using giant mechanical arms to secure the rocket on its return, in the same way as happened on the fifth test; on a floating platform in the Gulf of Mexico; in the Pacific Ocean near Hawaii; in the southeast Pacific; or in the Indian Ocean.Depending on the operational phase of the program, for landings at sea, both Starship and Super Heavy could have: (1) a hard landing at terminal velocity and break up on impact resulting in an explosive event at the surface of the water; (2) a soft water landing and tip over and sink or explode on impact at the surface of the water; and (3) breakup during reentry resulting in debris falling into the ocean, the FAA said in the document.The 120-meter-tall Starship has flown six times to date, with its flight cadence gradually increasing over time, except for a longer-than-expected interval between the fourth and fifth tests as SpaceX waited for a flight permit from the FAA.SpaceX is under pressure to ready a modified version of the Starship spacecraft for NASAs Artemis III mission, currently targeting September 2026, which will land the first woman and first person of color on the lunar surface.Editors Recommendations

Adani Group said its companies have enough cash to meet debt obligations, days after a U.S. indictment against founder Gautam Adani sparked a multibillion-dollar selloff in the Indian conglomerates stocks.

The EV makers Texas Gigafactory ran afoul of environmental overseers, with a faulty furnace door and hazardous wastewater.

Today moral rules apply to everything from the coffee we buy to the pronouns we use, and we are judged by the intensity of our commitment.

Fans clad in pink and green flocked to the musicals film adaptation. Gladiator II pulled in audiences eager to see Colosseum brawls.

Investigating delusions What delusions can tell us about the cognitive nature ofbelief Delusions can sometimes turn into strongly held beliefs. Michael Connors and Peter Halligan, The Conversation Nov 24, 2024 7:10 am | 25 Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreBeliefs are convictions of reality that we accept as true. They provide us with the basic mental scaffolding to understand and engage meaningfully in our world. Beliefs remain fundamental to our behavior and identity but are not well understood.Delusions, on the other hand, are fixed, usually false, beliefs that are strongly held but not widely shared. In previous work, we proposed that studying delusions provides unique insights into the cognitive nature of belief and its dysfunction.Based on evidence from delusions and other psychological disciplines, we offered a tentative five-stage cognitive model of belief formation.When faced with an unexpected sensory input or social communication, we seek to account for this based on existing beliefs, memories, and other social information. We then evaluate our account in terms of how well this explains our experiences and how consistent it is with our prior beliefs. If it passes these criteria, the belief is accepted. It then guides what we pay attention to and what other ideas we may consider.We propose that delusions can arise at different stages in this model. Our approach highlights the importance of the individuals search for meaning and social context in shaping delusions. It also draws attention to the impact of a delusion, once formed, on subsequent perceptions and thinking.This model linking delusions and beliefs differs from earlier accounts that suggested delusions were distinct from belief or arise as a largely passive response to anomalous sensory input such as a hallucination. Previous research, for example, has found that some people who believed that family members were replaced by impostors (known as Capgras delusion) had deficits in processing familiar faces, which could have generated this idea.Based on this, some have suggested that other delusions arise in a similar way but in combination with an as-yet undiscovered deficit in the cognitive process of evaluating our beliefs.But these accounts didnt fully consider other contributing factors, such as the individuals prior beliefs, social context, and their personal attempts to explain their experiences.Informative case studyThe study of delusions has been informed by select informative case studies. Unlike large group studies, case studies allow researchers a more detailed exploration of the origins and course of clinical features not explained by current theories.We recently published a paper in the international journal Cortex that describes a unique case study of a woman who temporally experienced compelling delusions during a brief hospital admission for postpartum psychosis, which can give rise to hallucinations, delusions, mood swings, and confusion. This is a rare complication of pregnancy, affecting around 12 in 1,000 women, thought to be due to hormonal changes or immunological factors.Natalie (a pseudonym) had no previous medical or psychiatric history. She developed postpartum psychosis while in hospital after the birth of her second child.As part of her condition, Natalie reported several delusions, including the belief that strangers were her parents-in-law in disguise (known as Fregoli delusion). Natalie recovered quickly with treatment. The combination of interviews and observations while she was experiencing the delusions and her later retrospective account offered a unique window into the onset and experience of her delusions.Following a full recovery, Natalie confirmed that she considered her delusions to be strongly held beliefs. She likened them to her conviction that her husband was her husband. This is contrary to some views that suggested that delusions are different from normal beliefs.Natalie was able to identify specific features that contributed to her delusions. In the case of believing that strangers were her in-laws, Natalie identified mannerisms, behaviors, and speech patterns of the strangers that reminded her of her in-laws. This suggested that the delusion could have arisen from inappropriate activation of memory representations of familiar people based on these cues and other factors.Natalie also recalled other beliefs, including that she was dead (known as Cotard delusion), which she did not share with clinicians at the time. She noted that she entertained this idea due to the failure of other explanations to account for her strange experiences and an idea from a television show.Natalie said she eventually dismissed this idea as implausible while still holding other delusional ideas. This suggests that belief evaluation may involve different thresholds for different delusions. It also highlights the private nature of some delusions.Across all of her delusions, Natalie described her active involvement in trying to explain and manage her experiences. She reported considering different explanations and testing these by seeking further information. For example, she asked questions of the people she thought were her in-laws. This suggests a surprisingly similar approach to how we typically form beliefs.Natalie recalled the influence of television and movies on her ideas. She also recalled how she elaborated on her delusions, once formed, based on information in her surroundings.These features challenge theories that delusions simply arise from anomalous sensory data. They instead highlight the role of the individuals search for meaning and social context, as well as the subsequent impact of delusions on perception and thinking.ImplicationsAs a case study, Natalies experiences are not necessarily representative of all people who experience delusions or postpartum psychosis. However, Natalies case presents informative features that theories of delusions need to account for.In particular, Natalies personalized insights highlight the critical role of the individual in actively trying to understand their experiences and bestow meaning. This is opposed to just passively accepting beliefs in response to anomalous sensory data or neuropsychological deficits. This suggests psychological therapies may be useful in treating psychosis, in combination with other treatments, in some cases.More generally, Natalies account reveals commonalities between delusions and ordinary beliefs and supports the view that delusions can be understood in terms of cognitive processes across the stages of normal belief formation that we identified.While there remain challenges in investigating delusions, further study may offer insights into the underpinnings of everyday belief and, in turn, of ourselves.Michael Connors, Conjoint Senior Lecturer in Psychiatry, UNSW Sydney, and Peter W Halligan, Hon Professor of Neuropsychology, Cardiff University. This article is republished from The Conversation under a Creative Commons license. Read the original article.Michael Connors and Peter Halligan, The Conversation The Conversation is an independent source of news and views, sourced from the academic and research community. Our team of editors work with these experts to share their knowledge with the wider public. Our aim is to allow for better understanding of current affairs and complex issues, and hopefully improve the quality of public discourse on them. 25 Comments

Go fish! Ancient fish-trapping network supported the rise of Maya civilization The Maya were landscape engineers on a grand scale, even when it came to fishing. Kiona N. Smith Nov 22, 2024 1:40 pm | 45 The Crooked Tree Wildlife Sanctuary is a biodiverse wetland in what's now Belizebut the precursors of the Maya turned it into an industrial-scale fishing operation. Credit: Fernando Flores The Crooked Tree Wildlife Sanctuary is a biodiverse wetland in what's now Belizebut the precursors of the Maya turned it into an industrial-scale fishing operation. Credit: Fernando Flores Story textSizeSmallStandardLargeWidth *StandardWideLinksStandardOrange* Subscribers only Learn moreOn the eve of the rise of the Maya civilization, people living in whats now Belize turned a whole wetland into a giant network of fish traps big enough to feed thousands of people.We already know that the Maya turned swamps into breadbaskets by draining and building raised blocks of land for maize fields. However, a recent survey of a wetland in whats now Belize suggests that the rise of the Maya civilization was fueled not just by maize but by tons of fish every year. University of New Hampshire archaeologist Eleanor Harrison-Buck and her colleagues recently mapped a network of channels and ponds for trapping fish, built just before the Maya civilization rose to prominence.Fish in a barrelHarrison-Buck and her fellow archeologists used drones and Google Earth data to map 108 kilometers of ancient channels that zigzag across 42 square kilometers of wetland in Belizes Crooked Tree Wildlife Sanctuary. The result is a network of channels and ponds that looks remarkably like the fish traps found farther south in Bolivia, built several centuries after the ones at Crooked Tree. Radiocarbon dating of material buried in the bottom of one channel suggests that the network has been around for at least 4,000 years.If Harrison-Buck and her colleagues are right, the channels, along with a series of ponds, are a system for trapping fish by channeling receding floodwaters into the ponds. In the ponds, harvesting the fish would have been easyjust like spearing fish in a barrel (or a small pond).The Crooked Tree Wildlife Sanctuary is a sprawling marshy grassland dotted with lakes and cut by streams. During the wet season, floods submerge the marshes and fish gather to spawn, as theyve done for thousands of years (at least). As the water recedes and the dry season sets in, retreating fish escape down the zigzag channels dug into the landscape, right into the ponds. And once the water levels drop even further, the fish in the ponds have no way to escape.Harrison-Buck and her colleagues calculated that at its peak, the system could have produced enough fish each year to feed around 15,000 people. Thats based on modern estimates of how many kilograms of fish people eat every year, combined with measurements of how many kilograms of fish people in Zambia harvest with similar traps. Of course, people at Crooked Tree would have needed to preserve the fish somehow, probably by salting, drying, or smoking them.Fisheries were more than capable of supporting year-round sedentarism and the emergence of complex society characteristic of Pre-Columbian Maya civilization in this area, write Harrison-Buck and her colleagues.When we think about the Maya economy, we think of canal networks and ditched or terraced fields. In just one patch of whats now Guatemala, a lidar survey revealed that Maya farmers drained thousands of acres of swampy wetland and built raised fields for maize, crossed by a grid of irrigation canals. To feed the ancient city of Holmul, the Maya turned a swamp into a breadbasket. But at least some of their precursors may have made it big on fish, not grain. The common feature, though, is an absolute lack of any chill whatsoever when it came to re-engineering whole landscapes to produce food. This Google Earth image shows the area containing the ancient fishery. Infrastructure built to last and lastFrom the ground, the channels that funneled fish into nearby ponds are nearly invisible today. But from above, especially during the dry season, they stand in stark contrast to the land around them, because vegetation grows rich and green in the moist soil at the base of the channels, even while everything around it dries up. That made aerial photography the perfect way to map them.In three of the channels, Harrison-Buck and her colleagues took samples of peat, which would once have lain in the bottom of the freshly dug passage, for radiocarbon testing. The results suggest that the channel had been there since at least 2000 BCE. Thats about 200 years before the start of the Formative Period, which marks the beginning of the rise of Maya civilization. (The Formative Period is also called the Preclassic Period; the period just before this is called the Archaic, so the channels were built very late in the Archaic Period.)The Maya evidently kept using the fish-trapping system even as their civilization grew and consolidated into large urban centers. Harrison-Buck and her colleagues found bits of Preclassic Maya pottery in some of the lower layers of sediment that had filled the channels, and other archaeologists have excavated once densely populated Maya communities along the shore of the Western Lagoon.Theres no sign that the Preclassic Maya did much to maintain the system of channels and ponds, though. Today, most of the channels have filled in with sediment, carried in by floodwaters, making them just subtle, curved dips in the ground, about 20 centimeters deep and 15 to 20 meters wide.But even now, the ancient channel system still works. While these features have filled in somewhat over the years, locals inform us that the ponds still concentrate fish during the dry season today, write Harrison-Buck and her colleagues.Science Advances, 2017. DOI: 10.1126/sciadv.adq1444 (About DOIs).Kiona N. SmithScience correspondentKiona N. SmithScience correspondent Kiona is a freelance science journalist and resident archaeology nerd at Ars Technica. 45 Comments