For thousands of years, a disease repeatedly struck ancient Eurasia, quickly spreading far and wide. The bite of infected fleas that lived on rats passed on the plague in its most infamous form — the Black Death of the 14th century — to humans, and remains its most common form of transmission today. During the Bronze Age, however, the plague bacterium, Yersinia pestis, had not yet developed the genetic tool kit that would allow later strains to be spread by fleas. Scientists have been baffled as to how the illness could have persisted at that time. Now, an international team of researchers has recovered the first ancient Yersinia pestis genome from a nonhuman host — a Bronze Age domesticated sheep that lived around 4,000 years ago in what is now modern-day Russia. The discovery has allowed the scientists to better understand the transmission and ecology of the disease in the ancient past, leading them to believe that livestock played a role in its spread throughout Eurasia. The findings were published Monday in the journal Cell. “Yersinia pestis is a zoonotic disease (transmitted between humans and animals) that emerged during prehistory, but so far the way that we have studied it using ancient DNA has been completely from human remains, which left us with a lot of questions and few answers about how humans were getting infected,” said lead author Ian Light-Maka, a doctoral researcher at the Max Planck Institute for Infection Biology in Berlin. There have been nearly 200 Y. pestis genomes recovered from ancient humans, the researchers wrote. Finding the ancient bacterium in an animal not only helps researchers understand how the bacterial lineage evolved, but it could also have implications for understanding modern diseases, Light-Maka added via email. “Evolution can sometimes be ‘lazy,’ finding the same type of solution independently for a similar problem — the genetic tools that worked for pestis to thrive for over 2000 years across over Eurasia might be reused again.” Unraveling the mystery of a Bronze Age plague The ancient bacterium that caused the Eurasia plague, known today as the Late Neolithic Bronze Age lineage, spread from Europe all the way to Mongolia, with evidence of the disease found across 6,000 kilometers (3,700 miles). Recent evidence suggests that the majority of modern human diseases emerged within the last 10,000 years and coincided with the domestication of animals such as livestock and pets, according to a release from the German research institute. Scientists suspected that animals other than rodents were a part of the enormous puzzle of the Bronze Age plague transmission, but without any bacterial genomes recovered from animal hosts, it was not clear which ones. To find the ancient plague genome, the study authors investigated Bronze Age animal remains from an archaeological site in Russia known as Arkaim. The settlement was once associated with a culture called Sintashta-Petrovka, known for its innovations in livestock. There, the researchers discovered the missing connection — the tooth of a 4,000-year-old sheep that was infected with the same plague bacteria found in humans from that area. Finding infected livestock suggests that the domesticated sheep served as a bridge between the humans and infected wild animals, said Dr. Taylor Hermes, a study coauthor and an assistant professor of anthropology at the University of Arkansas. “We’re sort of unveiling this in real time and trying to get a sense for how Bronze Age nomadic herders out in the Eurasian Steppe were setting the stage for disease transmission that potentially led to impacts elsewhere,” Hermes said, “not only in later in time, but also in a much more distant, distant landscape.” During this time within the Eurasian Steppe, as many as 20% of the bodies in some cemeteries are those of people who were infected with, and most likely died from, the plague, making it an extremely pervasive disease, Hermes said. While livestock are seemingly a part of what made the disease so widespread, they are only one piece of the puzzle. The identification of the bacterial lineage in an animal opens new avenues for researching this disease’s evolution as well as the later lineage that caused the Black Death in Europe and the plague that’s still around today, he added. “It’s not surprising, but it is VERY cool to see (the DNA) isolated from an ancient animal. It’s extremely difficult to find it in humans and even more so in animal remains, so this is really interesting and significant,” Hendrik Poinar, evolutionary geneticist and director of the Ancient DNA Centre at McMaster University in Hamilton, Ontario, wrote in an email. Poinar was not involved with the study. It is likely that humans and animals were passing the strains back and forth, but it isn’t clear how they did so — or how sheep were infected in the first place. It is possible sheep picked up the bacteria through a food or water source and then transmitted the disease to humans via the animal’s contaminated meat, he added. “I think it shows how extremely successful (if you want to label it that way) this particular pathogen has been,” Poinar added. He, as well as the study’s authors, said they hope that further research uncovers other animals infected with the ancient strain to further the understanding of the disease’s spread and evolution. Ancient plague to modern plague While the plague lineage that persisted during the Bronze Age is extinct, Yersinia pestis is still around in parts of Africa and Asia as well as the western United States, Brazil and Peru. But it’s rare to encounter the bacteria, with only 1,000 to 2,000 cases of plague annually worldwide. There is no need for alarm when it comes to dealing with livestock and pets, Hermes said. The findings are a reminder that animals carry diseases that are transmittable to humans. Be cautious when cooking meat, or to take care when bitten by an animal, he added. “The takeaway is that humans aren’t alone in disease, and this has been true for thousands of years. The ways we are drastically changing our environment and how wild and domesticated animals are connected to us have the potential to change how disease can come into our communities,” Light-Maka said. “And if you see a dead prairie dog, maybe don’t go and touch it.” Source of the article

OpenAI claims that its new flagship model, GPT-5, marks “a significant step along the path to AGI” – that is, the artificial general intelligence that AI bosses and self-proclaimed experts often claim is around the corner. According to OpenAI’s own definition, AGI would be “a highly autonomous system that outperforms humans at most economically valuable work”. Setting aside whether this is something humanity should be striving for, OpenAI CEO Sam Altman’s arguments for GPT-5 being a “significant step” in this direction sound remarkably unspectacular. He claims GPT-5 is better at writing computer code than its predecessors. It is said to “hallucinate” a bit less, and is a bit better at following instructions – especially when they require following multiple steps and using other software. The model is also apparently safer and less “sycophantic”, because it will not deceive the user or provide potentially harmful information just to please them. Altman does say that “GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD-level expert”. Yet it still doesn’t have a clue about whether anything it says is accurate, as you can see from its attempt below to draw a map of North America. It also cannot learn from its own experience, or achieve more than 42% accuracy on a challenging benchmark like “Humanity’s Last Exam”, which contains hard questions on all kinds of scientific (and other) subject matter. This is slightly below the 44% that Grok 4, the model recently released by Elon Musk’s xAI, is said to have achieved. The main technical innovation behind GPT-5 seems to be the introduction of a “router”. This decides which model of GPT to delegate to when asked a question, essentially asking itself how much effort to invest in computing its answers (then improving over time by learning from feedback about its previous choices). The options for delegation include the previous leading models of GPT and also a new “deeper reasoning” model called GPT-5 Thinking. It’s not clear what this new model actually is. OpenAI isn’t saying it is underpinned by any new algorithms or trained on any new data (since all available data was pretty much being used already). One might therefore speculate that this model is really just another way of controlling existing models with repeated queries and pushing them to work harder until it produces better results. What LLMs are It was back in 2017 when researchers at Google found out that a new type of AI architecture was capable of capturing tremendously complex patterns within long sequences of words that underpin the structure of human language. By training these so-called large language models (LLMs) on large amounts of text, they could respond to prompts from a user by mapping a sequence of words to its most likely continuation in accordance with the patterns present in the dataset. This approach to mimicking human intelligence became better and better as LLMs were trained on larger and larger amounts of data – leading to systems like ChatGPT. Ultimately, these models just encode a humongous table of stimuli and responses. A user prompt is the stimulus, and the model might just as well look it up in a table to determine the best response. Considering how simple this idea seems, it’s astounding that LLMs have eclipsed the capabilities of many other AI systems – if not in terms of accuracy and reliability, certainly in terms of flexibility and usability. The jury may still be out on whether these systems could ever be capable of true reasoning, or understanding the world in ways similar to ours, or keeping track of their experiences to refine their behaviour correctly – all arguably necessary ingredients of AGI. In the meantime, an industry of AI software companies has sprung up that focuses on “taming” general purpose LLMs to be more reliable and predictable for specific use cases. Having studied how to write the most effective prompts, their software might prompt a model multiple times, or use numerous LLMs, adjusting the instructions until it gets the desired result. In some cases, they might “fine-tune” an LLM with small-scale add-ons to make them more effective. OpenAI’s new router is in the same vein, except it’s built into GPT-5. If this move succeeds, the engineers of companies further down the AI supply chain will be needed less and less. GPT-5 would also be cheaper to users than its LLM competitors because it would be more useful without these embellishments. At the same time, this may well be an admission that we have reached a point where LLMs cannot be improved much further to deliver on the promise of AGI. If so, it will vindicate those scientists and industry experts who have been arguing for a while that it won’t be possible to overcome the current limitations in AI without moving beyond LLM architectures. Old wine into new models? OpenAI’s new emphasis on routing also harks back to the “meta reasoning” that gained prominence in AI in the 1990s, based on the idea of “reasoning about reasoning”. Imagine, for example, you were trying to calculate an optimal travel route on a complex map. Heading off in the right direction is easy, but every time you consider another 100 alternatives for the remainder of the route, you will likely only get an improvement of 5% on your previous best option. At every point of the journey, the question is how much more thinking it’s worth doing. This kind of reasoning is important for dealing with complex tasks by breaking them down into smaller problems that can be solved with more specialised components. This was the predominant paradigm in AI until the focus shifted to general-purpose LLMs. It is possible that the release of GPT-5 marks a shift in the evolution of AI which, even if it is not a return to this approach, might usher in the end of creating ever more complicated models whose thought processes are impossible for anyone to understand. Whether that could put us on a path toward AGI is hard to say. But it might create an opportunity to move towards creating AIs we can control using rigorous engineering methods. And it might help us remember that the original vision of AI was not only to replicate human intelligence, but also to better understand it. Source of the article

The Mayan Languages Preservation and Digitization Project promotes tools designed by and for Indigenous communities, like online glossaries and special phone keyboards Earlier this year, audiences at UNESCO’s Language Technologies for All conference witnessed a landmark event. For the first time in the international organization’s 79-year history, a native speaker of the Mayan language Kaqchikel delivered a keynote conference speech entirely in his ancestral tongue. When Maya educator Kawoq Baldomero Cuma Chávez stood on the podium at UNESCO’s Paris headquarters and started speaking in Kaqchikel, the audience—used to English and French as the conference’s lingua francas—paused in surprise. Cuma Chávez seized the opportunity, claiming a space to empower his language and those of other minority groups that have long been excluded from the global stage. After the presentation, delegates from around the world approached Cuma Chávez to share how much he’d inspired them. “They told me, ‘This is a good beginning for us to do the same with our languages,’” he recalls. For many, the name “Maya” prompts images of ancient pyramids and lost cities, remnants of a civilization that thrived in Mesoamerica until the Spanish conquest in the 16th century. But the Maya are not people of the past. Today, more than seven million Maya live in countries like Mexico, Guatemala, Belize, Honduras and El Salvador, maintaining distinct cultural practices, traditions and languages that they trace back to their ancestors. Central to Maya identity are the Mayan languages, a family of around 30 distinct but related languages still spoken today, including K’iche’, Yucatec Maya, Q’eqchi’, Mam and Kaqchikel. These languages have evolved over the centuries and remain a significant part of everyday life for modern Maya. In Guatemala, a country of roughly 18 million people, more than 40 percent of residents identify as Indigenous, and more than six million speak at least one of 22 Mayan languages. Cuma Chávez has dedicated more than two decades of his life to teaching Kaqchikel and preserving the Maya worldview through storytelling and poetry. To him, the keynote speech at the UNESCO conference wasn’t just symbolic. It was proof that Mayan languages, often confined to homes and villages in what was once called Mesoamerica, have every right to be heard in international arenas. “I am proud to be Indigenous,” Cuma Chávez says. “I have my language. I have my culture.” This philosophy of Indigenous pride drives the Mayan Languages Preservation and Digitization Project (MLPP), an open-source initiative launched in 2023 to preserve about 20 Mayan languages. The team behind the project builds free, easy-to-use tools—from digital glossaries to Android keyboards—that allow Mayan speakers to more easily communicate in their languages both online and offline. Through this work, MLPP is developing a step-by-step model that other marginalized language communities can follow to preserve their own linguistic heritage. Unlike language preservation efforts led by academic institutions and governments, which often focus on compiling scholarly archives that are difficult for the broader public to access, MLPP is a grassroots initiative designed by and for Mayan speakers. The tools the group creates are practical and owned by the community. According to UNESCO, 40 percent of people around the world don’t have access to education in a language they understand. This harsh reality traps entire communities in cycles of poverty, says Ludmila Golovine, CEO and president of the global translation and interpretation service MasterWord. “A person cannot learn a new language and a new subject at the same time,” she explains. Golovine is the great-great-niece of Yuri Knorozov, the Soviet linguist credited with decoding Maya glyphs in the mid-20th century. Her commitment to making languages accessible stems in part from her upbringing in the Soviet Union, where censorship was the norm. “Through language, you can actually gain access to knowledge,” Golovine says. “I see language as a step [toward] independence and freedom. Language access … should happen on the terms of the communities.” With support from MasterWord, MLPP has built a small team of full-time staff; contractors; and 100-plus volunteers living in Guatemala and beyond, including in the United States and Japan. The project blends longstanding cultural expertise with cutting-edge technology. At its core, it aims to empower communities and assert that Mayan languages belong in classrooms, courtrooms and cyberspace. “For the future, I wish [that] Mayan languages would have the same value as all others … that they would be as accepted as French, German or Italian,” says Cuma Chávez, who serves as an adviser to MLPP. Maya identity goes far beyond geographic location and blood ties to an ancient civilization. To call oneself Maya is to be rooted in community, sharing experiences, knowledge, gastronomy and other cultural elements, says Cuma Chávez. Walter E. Little, an anthropologist at the University at Albany who has worked with Maya communities for decades and contributes informally to MLPP, says the essence of Maya identity is found not in ancient monuments or genetic lineages, but rather in everyday practice. “A really important part of being Maya is practicing a particular type of culture,” he explains. Language, stories, dress and social ties define who you are. “The shared knowledge of all those things together—that’s what makes somebody Mesoamerican,” says Little, who is also the author of Mayas in the Marketplace: Tourism, Globalization and Cultural Identity. Outsiders who immerse themselves in these traditions, perhaps by learning a Mayan language or eating Maya food, might be jokingly called “Maya” by their friends, Little says—not because of their ancestry, but because culture is what matters most. This understanding of Maya identity stands in stark contrast to norms in the U.S., where many Native American tribes use “blood quantum,” a controversial measure of how much “Indian blood” one has, to determine enrollment eligibility. The system is a holdover from the late 19th and early 20th centuries, when the federal government embraced blood quantum as a way of limiting tribal citizenship. Despite the staggering linguistic diversity among modern-day Maya, several threads tie these distinct communities together: the milpa diet of maize, beans and squash; intricate weaving techniques; and communal storytelling traditions. These cultural anchors have survived centuries of upheaval. Beginning in the early 16th century, Spanish colonizers invaded Maya lands across Mesoamerica. The Spanish sought to conquer and impose Christianity upon the Maya, whose communities endured centuries of forced labor, displacement, and attempts to erase their languages and spiritual practices. The effects of this suppression remain visible today. “In Guatemala, discrimination for speaking [Mayan languages] is deeply ingrained,” Cuma Chávez says. The Spanish word “indio,” a derogatory term historically used to label Indigenous people as uncivilized, is still wielded as an insult, reflecting the lingering legacy of a colonial mindset that devalues Indigenous identity. While Western narratives often portray the Spanish conquest as an all-consuming tide of destruction, the Maya tell a different story: one of subtle, persistent resistance. “To speak of conquest is an overstatement,” Little says. “People maintained language, food cultures, weaving ... and learned tons of new things to their advantage.” Maya culture is inherently additive. New influences—whether from Spanish colonizers or Western technology—are not seen as replacements, but as layers added to existing traditions. Christianity and Maya spirituality, for example, coexist today: “The idea that ‘if you add something, you have to give up something’ is not really part of [Maya culture],” Little says. During the colonial period, some Maya scribes and leaders learned European writing systems, like the Latin alphabet introduced by Spanish missionaries, not as a means of assimilation, but rather to pass down stories and cultural knowledge in their native languages. Their efforts preserved enduring works of Maya literature, like the Popol Vuh and the Annals of the Cakchiquels. In other cases, Indigenous people forcibly relocated by the Spanish to centralized settlements known as reducciones, where they could be more easily controlled and converted, adopted the Mayan languages used by locals rather than switching to Spanish, Little says. These quiet acts of resistance ensured cultural survival in the face of erasure. But colonization still left scars. For generations, speaking a Mayan language has been seen as a barrier to opportunity, leading many Maya to internalize feelings of inferiority. Dulce María Horn, MLPP’s project coordinator, says the initiative seeks to reverse this mindset. “How do you start to decolonize yourself?” she asks. “It starts with saying, ‘I’m going to speak my language. I’m going to represent myself.’” MLPP began as a practical response to the language barriers Maya migrants face in legal and medical settings. In 2021, the U.S. was home to an estimated 1.8 million Guatemalan immigrants, many of them Maya. These individuals left their home country to escape discrimination and violence; they also sought economic opportunity. For individuals who don’t speak Spanish, only Mayan languages, finding interpreters to help navigate American immigration courts and hospital systems is a significant challenge. “How do you expect them to understand legal terminology in English?” Cuma Chávez asks. The absence of accurate interpretation can result in life-altering injustices, like being denied medical care due to misunderstandings about symptoms or getting in trouble for struggling to grasp complex legal matters. MLPP offers free, community-driven language tools to bridge these gaps, including training workshops where Indigenous volunteers learn to create and update digital resources themselves. In order to serve the people who need translation services and technology the most, “the community needs to always be in the driver’s seat,” says MLPP’s director, cultural anthropologist Winston Scott. Community meetings held in person in Guatemala or virtually with team members and volunteers around the world inform every decision made by MLPP, from which languages to prioritize to how glossaries should be recorded. “The project has grown into more of a ‘language access, language rights’ project,” Scott explains, “to ensure that all speakers … are able to access their language in meaningful ways.” One of MLPP’s earliest successes was the creation of six Mayan-language keyboards for Android devices. By adapting existing Spanish keyboard layouts to include essential Mayan characters, the team made it possible for users to communicate fluidly on apps like WhatsApp and Facebook without sacrificing their linguistic identity. The project’s online talking glossaries, a collection of written, visual and audio content in various Mayan languages, is regularly updated by volunteers who lend their voices and knowledge. These glossaries are at the heart of the team’s efforts. But MLPP isn’t stopping there. A partnership with a major tech company is in the works to develop text-to-speech services for K’iche’ and Q’eqchi’, allowing people who speak the languages but can’t read them to listen to written messages. In the long term, MLPP is supporting the development of the first neural machine translation model for Mayan languages. This artificial intelligence system will automatically translate text, operating similarly to Google Translate. But the model will be trained specifically on Mayan languages, avoiding issues raised by using Western source texts like the Bible, which can be filled with metaphors that don’t match everyday Mayan speech or the Maya’s values and worldview, says Scott. For MLPP, the preservation of Mayan languages isn’t just about archiving words. It’s about ensuring that young Indigenous people see their language as a living, evolving part of their future. This philosophy is embedded in the project’s partnership with the Bilingual Technology Center, a school located in Tucurú, a small town in Guatemala’s central highlands, where students learn Q’eqchi’, Spanish and English simultaneously. Pupils use MLPP keyboards to write poems, short stories and fables in their native Mayan tongue. Some of the students told Horn that this was their “first opportunity to use my language outside of my home.” According to Horn, the children added, “This is the first opportunity that I’ve ever had where somebody’s told me my language matters.” By showing students that their mother tongues can be used in digital storytelling, education and professional spaces, MLPP is dismantling the notion that Mayan languages are only for informal or domestic use. The project wants to demonstrate that speaking a Mayan language can open doors rather than close them, Horn says. Beyond the classroom, MLPP is collaborating with UNESCO to exchange resources and publish a practical guide to digitization, as well as with Cholsamaj, a Guatemala-based publishing house, to produce books in multiple Mayan languages. The goal is to provide engaging, age-appropriate reading materials that go beyond traditional dictionaries and grammar books. “If I’m a kid,” says Little, “I want to read comic books, graphic novels and cool stuff. I don’t want to read grammar.” MLPP seeks to rewrite the narrative that Mayan languages are outdated or inferior. “We believe that you can preserve the [Mayan] languages and learn Spanish as well … but we don’t think that you have to abandon the mother tongue to make that happen,” Scott says. Next, the team hopes to partner with Guatemala’s Department of Culture and Department of Education to publicize the tools developed by MLPP, Golovine says. MLPP envisions a future where Mayan languages have equal footing with other languages. “We want to see a Guatemala where everybody, no matter what their preferred language is, [can] feel proud of who they are,” Scott says. “We want to see … everybody have the opportunity to have an education in their language and to be able to choose their own path forward.” Source of the article