Latest news with #multilingual
Asharq Al-Awsat
2 days ago
- Business
- Asharq Al-Awsat
Microsoft Looks to Boost AI Performance in European Languages
US tech behemoth Microsoft is investing millions of dollars to funnel more European-language data into AI development, company president Brad Smith told AFP Monday. With today's leading AI models mostly trained on material in English, "the survival of these languages and the health of these cultures is quite literally at stake" without a course correction, Smith said in an interview. AI models are "less capable when it is in a language that has insufficient data," he added -- which could push more users to switch to English even when it is not their native language. Microsoft will from September set up research units in the eastern French city Strasbourg to "help expand the availability of multilingual data for AI development" in at least 10 of the European Union's 24 languages, including Estonian and Greek. The work will include digitizing books and recording hundreds of hours of audio. "This isn't about creating data for Microsoft to own. It is about creating data for the public to be able to use," Smith said, adding that the information would be shared on an open-source basis. The US-based company has in recent months striven to position itself as especially compatible with a gathering political push for European technological sovereignty. Leaders in the bloc have grown increasingly nervous at their dependency on US tech firms and infrastructure since Donald Trump's reelection to the White House. In June, Microsoft said it was stepping up cooperation with European governments on cybersecurity and announced new "data sovereignty" measures for its data centers on the continent. Smith said that Monday's announcement was just the latest evidence of the company's commitment to Europe. Most leading AI firms are American or Chinese, although Europe has some standouts like France's Mistral or Franco-American platform Hugging Face. Away from Microsoft, some European initiatives such as TildeLM are pushing to develop local-language AI models. The Windows and Office developer also said Monday that it was working on a digital recreation of Paris' Notre-Dame cathedral that it plans to gift to the French state, as well as digitizing items from the country's BNF national library and Decorative Arts Museum.
Geeky Gadgets
2 days ago
- Science
- Geeky Gadgets
Kimi K2 Agent Researcher for Deep Reasoning Research Tasks
What if you could delegate your most complex research tasks to an AI that not only understands the intricacies of your work but also evolves with every challenge it faces? Enter the Kimi K2 Agent Researcher, a new single-agent system designed to redefine how we approach deep reasoning and long-term problem-solving. Unlike traditional tools that falter under the weight of extended tasks or lose focus in the noise of irrelevant data, the Kimi K2 thrives in complexity, offering precision, adaptability, and unparalleled efficiency. Imagine a system that can sift through hundreds of sources, refine hypotheses on the fly, and deliver actionable insights—all while maintaining a laser-sharp focus on your objectives. It's not just a tool; it's a partner in innovation. Prompt Engineering explores the fantastic potential of the Kimi K2 Agent Researcher, delving into its innovative functionalities like iterative hypothesis refinement, real-time internal search, and automated coding. You'll discover how its single-agent architecture eliminates inefficiencies common in multi-agent systems, making sure consistency and clarity even in the most demanding research environments. Whether you're a data scientist navigating complex datasets or an academic pushing the boundaries of your field, the Kimi K2 promises to elevate your research process. But how does it compare to other AI models, and what makes its design uniquely suited for global, multilingual challenges? The answers lie in its seamless integration of technology and purpose—a design philosophy that might just change the way we think about research forever. Kimi K2 Research Overview Core Features of the Kimi K2 Agent Researcher At the foundation of the Kimi K2 Agent Researcher lies its ability to handle complex research tasks with exceptional accuracy and efficiency. Its single-agent architecture incorporates three primary tools that work in tandem to optimize performance: Real-time internal search: This feature enables the system to swiftly retrieve relevant information from internal datasets, making sure rapid access to critical data and minimizing delays in research workflows. This feature enables the system to swiftly retrieve relevant information from internal datasets, making sure rapid access to critical data and minimizing delays in research workflows. Text-based browser: Equipped to conduct extensive web-based research, this tool can explore up to 200 URLs per task, allowing comprehensive data collection from diverse online sources. Equipped to conduct extensive web-based research, this tool can explore up to 200 URLs per task, allowing comprehensive data collection from diverse online sources. Automated coding tool: Designed to generate and refine code, this tool supports technical aspects of research, streamlining processes that would otherwise require significant manual effort. By combining these tools, the system synthesizes information from multiple sources, delivering thorough analyses and highly accurate results. This integration ensures that users can rely on the system for both breadth and depth in their research endeavors. Training Methodology and Advanced Functionalities The Kimi K2 Agent Researcher is trained using an end-to-end reinforcement learning approach, allowing it to refine its strategies through iterative trial and error. This training methodology underpins several advanced functionalities that set the system apart: Iterative hypothesis refinement: The system evaluates conflicting information, adjusts hypotheses, and self-corrects to enhance the accuracy of its conclusions. The system evaluates conflicting information, adjusts hypotheses, and self-corrects to enhance the accuracy of its conclusions. Information validation: It verifies the reliability and accuracy of data before presenting results, making sure that conclusions are based on credible sources. It verifies the reliability and accuracy of data before presenting results, making sure that conclusions are based on credible sources. Context management: By retaining relevant information and filtering out irrelevant data, the system maintains clarity and focus during extended research tasks. These capabilities make the Kimi K2 Agent Researcher particularly effective for scenarios requiring deep reasoning, such as scientific research, data analysis, and solving complex problems. Its ability to adapt and refine its approach ensures consistent performance, even in dynamic or uncertain research environments. Kimi K2 Agent Researcher Watch this video on YouTube. Take a look at other insightful guides from our broad collection that might capture your interest in Kimi K2 . Performance Metrics and Comparative Benchmarks The Kimi K2 Agent Researcher delivers impressive performance metrics, particularly in tasks requiring deep reasoning and extended focus. It can execute up to 23 reasoning steps within a single task and supports up to 50 iterations without experiencing 'context rot,' a common issue in prolonged tasks where systems lose track of relevant information. This resilience ensures that the system maintains accuracy and coherence, even in demanding scenarios. While the Kimi K2 Agent Researcher outperforms most comparable models in terms of versatility and integration, it does fall slightly behind the Gro 4 model in specific benchmarks. However, its ability to incorporate diverse data sources, including Chinese web links, gives it a distinct advantage for global research applications. This feature broadens its utility for users who require access to multilingual or region-specific data. Single-Agent Design and Its Advantages Unlike multi-agent systems, which distribute tasks among specialized agents, the Kimi K2 Agent Researcher employs a holistic single-agent design. This approach simplifies coordination and enhances the system's ability to manage large observation contexts. By focusing on a unified problem-solving strategy, the system reduces redundancy and ensures a streamlined research process. The single-agent architecture also allows for greater consistency in reasoning and decision-making. It eliminates the potential for miscommunication or inefficiencies that can arise in multi-agent setups, making it particularly well-suited for tasks that require sustained focus and comprehensive analysis. API Hosting Options and User Accessibility The Kimi K2 Agent Researcher offers flexible API hosting options, allowing users to select configurations that align with their specific needs and budgets. These options include variations in quantization levels, token processing speeds, and pricing structures, making sure that the system can accommodate a wide range of use cases. Beyond its technical capabilities, the system enhances user accessibility through its reporting and visualization features. It generates detailed reports and interactive websites to summarize findings, simplifying the interpretation and application of results. This functionality is particularly valuable for professionals who need to present their research in a clear and actionable format. Additionally, the system provides a limited number of free searches per month, allowing users to explore its capabilities before committing to a subscription. Its balanced interaction style ensures that information is delivered accurately and without unnecessary bias, fostering a productive and engaging research experience. Why the Kimi K2 Agent Researcher Stands Out The Kimi K2 Agent Researcher distinguishes itself as a powerful tool for addressing complex research challenges. Its advanced reasoning capabilities, rigorous validation processes, and robust context management make it a reliable choice for professionals seeking precision and adaptability. Whether you are conducting academic research, analyzing large datasets, or exploring new hypotheses, this single-agent system provides the tools and efficiency necessary to achieve your objectives with confidence. By combining innovative technology with user-centric design, the Kimi K2 Agent Researcher offers a comprehensive solution for modern research needs. Its ability to integrate diverse data sources, adapt to evolving tasks, and deliver actionable insights ensures that it remains a valuable resource for professionals across industries. Media Credit: Prompt Engineering Filed Under: AI, Top News Latest Geeky Gadgets Deals Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, Geeky Gadgets may earn an affiliate commission. Learn about our Disclosure Policy.
New York Times
5 days ago
- General
- New York Times
‘Bonjour' Sets Off a Linguistic Dispute on a Belgian Train
A train attendant's use of both Dutch and French — 'goeiemorgen' and 'bonjour' — to greet passengers during rush hour last year could easily have gone unnoticed in multilingual Belgium. But it rubbed one Dutch-speaking commuter the wrong way. He told the attendant, llyass Alba, that 'bonjour' was not appropriate because they were still in Dutch-speaking Flanders. 'Excusez-moi?' Mr. Alba replied. 'You need to speak in Dutch,' the man said back to him in Dutch. The dispute escalated. Mr. Alba explained what was going on to nearby passengers — in French. The Dutch speaker then filed a complaint to Belgium's Permanent Commission for Linguistic Control. Five months after the initial incident, in March, the commission issued its opinion: The commuter's complaint was 'well founded.' Their guidance was reported on Friday by the Dutch-language newspaper Nieuwsblad. Invoking administrative language laws, the Permanent Commission for Linguistic Control stated that the train attendant should have greeted passengers in Dutch. 'As soon as he knows the traveler's language (French or Dutch), he will respond in that language (French or Dutch),' the commission said in a statement. Want all of The Times? Subscribe.
Entrepreneur
6 days ago
- Business
- Entrepreneur
United We Care Launches Shunya Labs to Revolutionise AI Voice Infrastructure
The platform supports over 32 Indic languages, including Hindi, Marathi, Assamese, and Maithili. Seven additional languages are currently being integrated. You're reading Entrepreneur India, an international franchise of Entrepreneur Media. In a bold move to redefine the future of voice technology, mental health AI startup United We Care has unveiled Shunya Labs, a deep tech initiative focused on building advanced AI speech infrastructure. Positioned as a game-changer in automatic speech recognition (ASR), Shunya Labs aims to set a new standard in accuracy, speed, and multilingual capability, while maintaining a strong emphasis on privacy and transparency. According to the company, the platform is equipped with an industry-leading ASR engine and real-time, on-premise processing capabilities. Shunya Labs is operational and already demonstrating potential to outperform Big Tech leaders across critical performance metrics. Designed with India's linguistic diversity at its core, the platform supports over 32 Indic languages, including Hindi, Marathi, Assamese, and Maithili. Seven additional languages are currently being integrated. Its use cases span from rural telehealth kiosks to high-demand environments like call centres and defence applications. "We didn't set out to beat the benchmarks — we set out to invent what didn't exist," said Ritu Mehrotra, Founder of United We Care. "And when we built it, we realized we'd created something the world didn't know it needed: AI voice infrastructure that listens like a human, runs like a machine, and respects the sanctity of privacy." Shunya Labs is engineered for efficiency, promising to reduce enterprise cloud expenses by a factor of twenty. It also supports edge deployment, which enables offline functionality and faster inference times, especially critical in remote or sensitive environments. The technology behind Shunya Labs has already shown its capabilities through its integration with Stella, United We Care's AI wellness engine. Stella's foundation in Shunya's ASR technology has led to the creation of a Clinical Knowledge Graph comprising over 230 million nodes and the development of the Spatio-Temporal Graph Attention Network (STGAT), both of which highlight the potential for emotional intelligence and clinical reasoning in AI. "Shunya isn't just a name. It's our origin point — a nod to the Indian discovery of zero that changed mathematics forever," said Sourav Banerjee, Co-founder and CTO. "At Shunya Labs, we start from first principles, engineer with surgical precision, and build for the future — not features. In a world drowning in noise, we build the intelligence layer that actually listens." With plans underway for deployment across multiple continents and an upcoming release on Hugging Face, Shunya Labs is positioning itself not merely as an interface for voice AI, but as the backbone of future-ready infrastructure.
Daily Mail
12-07-2025
- General
- Daily Mail
EXCLUSIVE Revealed: The 2,000 schools where more than HALF of pupils don't speak English as their first language... so is YOUR child's one of them?
English is no longer the first language for the majority of pupils at more than 2,000 schools, MailOnline can today reveal. No children at two primary schools – one in Tower Hamlets and another in Kirklees, West Yorkshire – have English as their mother tongue. Our statistics, obtained exclusively under Freedom of Information (FOI) laws, show nine in ten pupils don't speak only English at home at 107 schools. The full results of our audit, covering all 22,000 state schools, can be viewed in our search tool below. Bengali is the mother tongue of 92 per cent of the pupils at Kobi Nazrul, the primary school in Tower Hamlets where none of the kids solely speak English at home. The others speak a slew of other languages, including Indonesian and Urdu. At Pentland Infant in Dewsbury – the Kirklees school – the overwhelming majority of the children speak either Gujarati (36 per cent) or Panjabi (45 per cent). When broken down by languages, our FOI revealed Tottenhall Infant School, Enfield had the highest rate of Albanian speakers at 18 per cent. Sheffield's Netherthorpe Primary School topped the league table for Arabic (54 per cent), meanwhile Burnley Brow Community School in Oldham had the highest share of Bengali speakers (93 per cent). In terms of Chinese, St Cecilia's CofE Primary School in Wokingham, Berkshire, had the highest rate (28 per cent). Polish speakers were most heavily concentrated at St Cuthbert's Catholic Primary School in Windermere (43 per cent), while Urdu was most common at St Michael's CofE Primary School in Bolton (58 per cent). In total, English is no longer the first language for the majority of children at 2,039 schools. This includes deaf schools, where pupils' first language is BSL. Nationwide, English isn't the first language of 1.8 million pupils, or one in five pupils. Fuelled by immigration having spiralled to all-time highs, this has risen from 1.1 million ten years ago. The soaring numbers have raised concerns among critics that the slew of different languages can be incredibly disruptive in classrooms. Robert Bates, of the Centre for Migration Control, told MailOnline: 'It's a great shame that we have reached this point and it is evidence of successive governments failing to properly address assimilation issues. 'Multi-language classrooms are hugely problematic. They inevitably sap the precious time of teachers who are forced to devote additional attention to those with a weaker grasp of the language at the expense of those who only speak English. 'The resources used will be lowest-common-denominator teaching devices that provide little from the individuals to excel. 'They also foster long term division, with little incentive for those who do not speak English to improve their fluency. 'It is time for a sink or swim approach in which state funding for translation costs are stripped across the board, encouraging adult migrants and parents alike to devote greater attention to improving this pretty basic skill.' Schools have had to pivot their limited resources to provide for the needs of pupils who struggle with English. Strain has been placed on their tight budgets due to providing translated versions of resources, adding subtitling and voiceovers, as well funding in-class interpreters. Teachers have previously said schools were under mounting pressure from mass immigration and called on ministers to fund them properly to cope with the array of different languages that pupils speak. MailOnline's analysis showed the authority that had the highest rate of first language speakers other than English was Newham (66.4 per cent). Brent (63.7 per cent) and Harrow (63.6 per cent) rounded out the top three. On the other hand, our FOI found 97 per cent of children spoke English as their first language in Northumberland. Redcar and Cleveland (96.7 per cent) and Cornwall (96.1 per cent) came second and third, respectively. Experts warn that kids who aren't taught English before they get to school are less likely to do well. Ian Mansfield, head of education at Policy Exchange, said: 'These statistics demonstrate the very real pressure that mass immigration places on public services.' However, some studies suggest that pupils who speak English as a second language can outperform native speakers and their presence has no impact on the learning of other pupils. Despite the language barrier, Kobi Nazrul was regarded as 'friendly and welcoming' in a glowing report from Ofsted in its most recent inspection Depending on how close they are to English, both phonetically and grammatically, some languages can be very difficult to switch from. Alp Mehmet, of Migration Watch UK, said: 'English is the glue that brings and holds us together as a country and helps new arrivals to blend in.' Mr Mehmet, who said he was the only non-English speaker at his east London school growing up in the 1960s, added: 'If teachers have to focus their attention on children who struggle to speak English because the language spoken at home is different, the time devoted to other pupils is bound to diminish.' Jim McConalogue, CEO at the think-tank Civitas said: 'You are effectively seeing the proportion of school-age pupils in England speaking English as an additional language continuing to steadily increase over time. 'Measures to ensure English language acquisition must therefore be prioritised. 'There are some short-term challenges that the DoE need to address for students who speak English as an additional language, as a lack of emphasis on English language provision can have a detrimental impact on both these students themselves and their classmates. 'Many policymakers talk about diversity and inclusion, but relatively little is being done to support English language acquisition. 'This omission creates deeper problems not only for the individual children but society at large, around a common citizenship, a shared language and identity, belonging and social cohesion.' Some politicians believe the high number of immigrants now entering the UK need to be integrated into society by the government properly. Shadow education minister Neil O'Brien said: 'These figures underline just how serious the challenge of integration has become in some parts of the country. 'Integration shouldn't be an afterthought, it must be a priority. 'The government has totally failed to stop the boats and numbers coming illegally are massively up. 'They abandoned our plans to raise income requirements on the family route. And as well as having no plan on immigration they have no plan on integration.' Some campaigners argue the Government needs to have more involvement in schools that have especially high numbers of foreign language speakers. Chris McGovern, of the Campaign for Real Education, said: 'The Government should publish the data regarding the language of pupils in each school and ensure that non-native speakers do not outnumber native speakers of English. 'This will benefit all pupils and allow our society to become more integrated, harmonious and at peace with itself.' The FOI data, from the Department of Education, reflects the state-of-play in January 2024. In it, schools have some flexibility about how they code pupils' language. DoE bosses define the first language as 'where the pupil has been exposed to a language other than English during early development and continues to be exposed to this language in the home or in the community'. A Government spokesperson said: 'This government is determined to break down barriers to opportunity so every child, including pupils who speak a language other than English, can achieve and thrive. 'Schools are best placed to understand the needs of their pupils with English as an additional language and are responsible for determining what support to put in place. 'Additional funding is provided to schools to support pupils who need help learning how to speak, read and write English. 'More widely, to better integrate migrants into their new communities this government will introduce a new English language requirement across a broader range of immigration routes.' MailOnline's investigation comes after Keir Starmer last month warned that mass immigration risks making Britain an 'island of strangers'. Scrambling to blunt the threat of Reform, the Prime Minister vowed to give Brits what they had 'asked for time and time again'. He unveiled a package to 'take back control of our borders'. The skills threshold will be hiked and rules on fluency in English toughened under the Government's plan to bring down annual inflows by around 100,000.
