Latest news with #Llama3.1
The Wire
5 days ago
- Business
- The Wire
Two Systems, Two Spheres: The Slow, Painful Divorce of the US and China
Menu हिंदी తెలుగు اردو Home Politics Economy World Security Law Science Society Culture Editor's Pick Opinion Support independent journalism. Donate Now Top Stories Two Systems, Two Spheres: The Slow, Painful Divorce of the US and China Manoj Joshi 6 minutes ago Given the level of interdependence, the process is likely to be slow but will manifest itself through higher costs, slower innovation and geopolitical tensions. US President Donald Trump and Chinese President Xi Jinping. Photo: X/@Maks_NAFO_FELLA Real journalism holds power accountable Since 2015, The Wire has done just that. But we can continue only with your support. Contribute now The US-China negotiations on a trade deal have stalled for the moment. According to reports, the US is angry that while it came up with a 90-day tariff truce at the start of the negotiations, China has not reciprocated, as it had promised to resume exports of rare earth minerals. Last week's tirade by President Trump – accusing China of violating the interim agreement which he had provided to 'save' China from a bad economic situation – is a pointer to the intensity of the negotiations which could well be standing on the brink of a breakdown. Even if a deal is struck, it is clear that a gradual decoupling of the United States and China is well underway, encouraged by the policies that began in the Trump-I administration, carried on by President Biden and now being adopted by Trump II. This comes at a time when China has established its capabilities as a near-peer competitor of the US in a range of areas. In recent months, we have seen China testing two or maybe three sixth generation fighter jets, less than a decade after the US did so. We have seen the arrival of Deep Seek, an AI large language model with significantly lower training costs as compared to its peers like Open AI's GPT-4. It also used one-tenth the computing power consumed by Meta's Llama 3.1. Last week, the news portal Axios reported that China was now 'setting the pace in life sciences R&D', conducting more clinical trials and licensing new drugs as compared to US companies. This comes at a time when Trump is gutting the National Institutes of Health and bio-medical research in US universities. The news from China on the technology front has been generally upbeat. In October 2024, Bloomberg News, in a special analysis, said that of the 13 key technologies tracked by their researchers, the Chinese had a world leadership level in five of them and was catching up fast in seven others. The first list included UAVs, solar panels, graphene, high speed rail, electrical vehicles and batteries. In the second list were semiconductors, AI, robots, machine tools, large tractors, drugs and LNG carriers. The one area where the Chinese lagged was commercial aircrafts. In June 2024, the Economist cited the Leiden Ranking of the volume of scientific research output, there are now six Chinese universities or institutions in the world top ten. The Nature Index put the number at seven. Last week, the US announced that it would revoke visas for members of the Communist Party of China or for those studying in 'critical fields.' The administration has also announced a general policy of vetting future student visas, including their social media posts. This could see a sharp decline in foreign students, especially the Chinese, in the coming years. As of 2024, there were some 277,000 Chinese students in the US. The US alleges that China uses US universities to advance its military and technological capabilities. The number of US students in China has already declined sharply from some 11,000 in 2019 to just 1,000 in 2024. Trump's National Security Strategy of 2017 first identified China as a strategic competitor rather than a partner. The Biden NSS built on this notion and its NSS of 2022 declared 'The People's Republic of China harbours the intention and, increasingly, the capacity to reshape the international order in favour of one that tilts the global playing field to its benefit.' It was the first Trump administration that launched a tariff war on China that morphed into a campaign of technology denial, a strategy that intensified in the Biden years. Trump II took office with a promise to hit China with the highest tariff rates. After reaching the absurd point of 145% American tariffs on Chinese goods, and a reciprocal 125% on American products, the two countries are currently trying to work out more reasonable trade arrangements though the prognosis is not good. But it is clear that the US policy of denial of certain high technologies, especially those relating to semiconductors and jet engines, as well as Chinese restrictions on some exports to the US will stay. In March, the Trump administration added over 80 companies, mainly Chinese, to the American blacklist making it tougher for US high-tech companies like Nvidia to do business with China. For its part, China has long sought an autonomy of sorts with regard to technological dependence on the West. It has strategically promoted certain industries through its policies which range from subsidie to demands for forced technology transfer and outright espionage. In addition, through its 'Thousand Talents Programme' it has obtained the services of scholars from foreign universities, including overseas Chinese, to push its own domestic research and development in critical areas. The Make in China initiative of Xi Jinping that began in 2015 sought to have China dominating in 10 advanced industries that would have a 70% domestic market share by this year. In addition to robotics, it included advanced rail equipment, high tech maritime vessels, aerospace and aviation equipment, electrical vehicles and information technology. Not surprisingly, this policy was strenuously opposed by western countries which had dominated the Chinese and global markets in these areas. According to the Financial Times, the Chinese used a unique combination of industrial policy, subsidies and other state support coupled with private sector entrepreneurism and 'ferocious competition in China's vast market' to match and, in some cases, best foreign competitors. It was the Chinese dominance of the industrial supply chains and exports that persuaded Trump to negotiate, rather than fight with China on trade last month. China today commands a full 29% of global manufacturing by value. Equally importantly, it controls critical supply lines for a range of key products. The FT report also indicates that China has some way to go in achieving all four of its goals – reducing import dependence, cutting reliance on foreign companies, becoming the technological leader and achieving global competitiveness. It has done so only in high speed rail and electric power equipment. It has displayed a strong performance in robots, machine tools, agricultural machinery, electrical vehicles, aerospace equipment and biomedicine. The one area in which China is weak is in the manufacture of commercial aircrafts. This is the reason the US has now decided to embargo supplies to the Chinese Comac C919, a commercial jet that is similar to the Airbus 320 and which entered service in 2023. The aircraft depends on western engines as well as other major components. The Trump administration's policies appear to have chosen the worst path available to them. Instead of gathering an alliance of partners to confront China's trade practices, the US alienated most of them by targeting them first. Trump enunciated a policy of calling on industry to manufacture in the United States, yet, through his policies, he has sought to gut the US R&D establishment and its famed universities which were the source of the intellectual capital that has underwritten US technological dominance. His attacks on elite US universities are likely to drive talent to other countries, especially China. Whatever be the case, the world seems headed for a decoupling of the two great powers, a policy that will result in the creation of two techno-spheres. This will be a messy and costly process and will inevitably extract a price from the global economy when countries have to confront two technology standards and two antagonistic supply chains. Even so, given the level of interdependence of the two economies, the process is likely to be slow but manifest itself in the coming years through higher costs, slower innovation, supply chain disruptions and geopolitical tensions. Manoj Joshi is a distinguished fellow with the Observer Research Foundation in Delhi. This piece was first published on The India Cable – a premium newsletter from The Wire & Galileo Ideas – and has been updated and republished here. To subscribe to The India Cable, click here. The Wire is now on WhatsApp. Follow our channel for sharp analysis and opinions on the latest developments. Make a contribution to Independent Journalism Related News US to Impose Visa Restrictions on Foreign Officials Accused of Censoring Americans Abroad India and China: Two Contrasting Models of Dealing With Trump's US Has Trump 2.0 Deprioritised India? The Evidence is Clear. As US Steps Back From Tariff War With China, What You Need to Know Harvard's Indian Students Are Trapped in Trump's Visa Crossfire US Targets Indian Travel Agents with Visa Bans as Part of Immigration Policy Trump Terms US-China Tariff Talks in Geneva a 'Very Good Meeting', Says Negotiated 'Total Reset' Trump Admin Pauses New Student Visas as it Mulls More Social Media Vetting US Cites National Security Grounds, Procedural Errors to Reject India's Notice at WTO About Us Contact Us Support Us © Copyright. All Rights Reserved.
Business Times
05-05-2025
- Business
- Business Times
Singapore's AI large language model Sea-Lion to offer more features as more firms use it in S-E Asia
[SINGAPORE] Singapore's home-grown large language model (LLM), Sea-Lion, is steadily gaining traction, with some 235,000 downloads so far, bolstered by adoption by large companies such as GoTo Group in Indonesia. After releasing the latest model with 'reasoning' capabilities on Apr 15, its researchers at AI Singapore told The Straits Times that they plan to add voice recognition later in 2025, followed by other modalities such as visual recognition. The new features are expected to enhance the model's usability in a region rich in spoken and unwritten languages. The model currently recognises 13 languages, including Javanese, Sudanese, Malay, Tamil, Thai and Vietnamese, as well as English and Chinese. Sea-Lion is already tapped by some businesses for its language features, with GoTo among the first enterprises to adopt Sea-Lion in February 2024 as a base to build its own artificial intelligence (AI) system. Its chief data officer, Ofir Shalev, noted: 'Training a model from scratch is often prohibitively expensive. So like many in the industry, we adopted a continuous pre-training approach – building on an existing model as the starting point.' GoTo's Sahabat-AI model is now benchmarked as more accurate in reading and interpreting Bahasa Indonesia, Javanese and Sundanese than other models of similar size, according to Shalev. A NEWSLETTER FOR YOU Friday, 8.30 am Asean Business Business insights centering on South-east Asia's fast-growing economies. Sign Up Sign Up The S$70 million Sea-Lion initiative to build an open-source LLM that reflects the native characteristics of South-east Asia was publicly launched in December 2023. It is funded by the National Research Foundation and backed by the Infocomm Media Development Authority and Agency for Science, Technology and Research. Sea-Lion's latest iteration v3.5, built on another open-source model, Llama 3.1, is fine-tuned to be more adept at complex problem-solving, logical inference and multi-step instructions than earlier versions. It is one of the few models worldwide to offer a 'hybrid reasoning' mode, allowing users to toggle advanced reasoning on or off – saving time and computing resources for straightforward tasks. A key improvement is the model's 128,000-token context window, enabling it to process and understand much longer documents and conversations without losing track of earlier information. This is on a par with leading models such as GPT-4o and Meta's Llama 3.1, and surpassed by only a few, such as Google's Gemini and Anthropic's Claude. The chief scientist at Singtel technology subsidiary NCS, Ying Shaowei, calls Sea-Lion v3.5 a significant advancement for Singapore's AI ecosystem. NCS is expanding its Sea-Lion pilot to support legal and compliance document translation. It will use the latest model to also perform multilingual customer engagement and cross-border regulatory change detection, and to unify internal content across languages and formats. Ying said: 'We are deeply interested in how well the model performs as part of a larger system, how it integrates with existing enterprise workflows, its interoperability, its total cost of deployment and its security posture.' Dr Leslie Teo (centre), senior director of AI products and the Sea-Lion team's lead, with the rest of the team at AI Singapore. The model currently recognises 13 languages, including Javanese, Sudanese, Malay, Tamil, Thai and Vietnamese, as well as English and Chinese. PHOTO: HESTER TAN, ST These critical factors will determine if Sea-Lion can be scaled up in real operational environments, Ying said. He added: 'Sea-Lion is showing real promise on several of these fronts, and we are continuing to test its role in more complex AI-driven solutions that go beyond language to include insight extraction, document classification and voice-based interaction.' In Thailand, Sea-Lion has been adopted into a voice app that, in one instance, helped a Bahasa Indonesia-speaking worker file a complaint with the Labour Department to recoup her unpaid salary. The complaint was filed in Thai. Its other use cases include a Python programming script that recognises the unique Thai calendar system of adding 543 years to the Gregorian year. It also recommended Asian condiments for cooking to a Filipino-speaking user. Dr Ngui Jian Gang, who demonstrated the model to The Straits Times, said: 'GPT 4 recommends mayonnaise, which is not very local, and also lemon butter sauce, which is delicious, but also not very local.' Dr Leslie Teo, senior director of AI products and the Sea-Lion team's lead at AI Singapore, said when evaluated against an industry benchmark tailored for the region, Sea-Lion v3.5 outperformed ChatGPT's and Deepseek's recent models. The benchmark called SEA-Helm – the South-east Asian Holistic Evaluation of Language Models – is developed by AI Singapore in partnership with Stanford University's Centre for Research on Foundation Models. The evaluation was done across five metrics – comprising natural language processing, instruction-following, conversational ability, linguistic and cultural performance, and toxicity detection for low-resource South-east Asian languages. Dr Teo describes Sea-Lion as best used as a 'small model' for simple tasks, or as a 'companion model' paired with large models such as ChatGPT, Claude or DeepSeek to fill gaps in the South-east Asian context. He hopes that it will be useful for organisations with operations in the region. With the new and planned improvements, he is hopeful of drawing more adopters. He said: 'What has changed this time is that performance is close to the frontier.' The Sea-Lion ecosystem is also ready, he added. An interactive web platform called Playground lets users try out the model; its Telegram bot allows users to engage it in their preferred language; and it has application programming interfaces that enable developers and organisations to integrate Sea-Lion into their applications and workflows. Dr Teo said: 'We feel that the model is good enough. We want people to use it more. We want to get to a point where we have real big users using it, criticising it and helping us make it better.' THE STRAITS TIMES
Reuters
02-04-2025
- Business
- Reuters
New AI benchmarks test speed of running AI applications
SAN FRANCISCO, April 2 (Reuters) - Artificial intelligence group MLCommons unveiled two new benchmarks that it said can help determine how quickly top-of-the-line hardware and software can run AI applications. Since the launch of OpenAI's ChatGPT over two years ago, chip companies have begun to shift their focus to making hardware that can efficiently run the code that allows millions of people to use AI tools. As the underlying models must respond to many more queries to power AI applications such as chatbots and search engines, MLCommons developed two new versions of its MLPerf benchmarks to gauge speed. One of the new benchmarks is based on Meta's (META.O), opens new tab so-called Llama 3.1 405-billion-parameter AI model, and the test targets general question answering, math and code generation. The new format tests a system's ability to process large queries and synthesize data from multiple sources. Nvidia (NVDA.O), opens new tab submitted several of its chips for the benchmark, and so did system builders such as Dell Technologies (DELL.N), opens new tab. There were no Advanced Micro Devices (AMD.O), opens new tab submissions for the large 405-billion-parameter benchmark, according to data provided by MLCommons. For the new test, Nvidia's latest generation of artificial intelligence servers - called Grace Blackwell, which have 72 Nvidia graphics processing units (GPUs) inside - was 2.8 to 3.4 times faster than the previous generation, even when only using eight GPUs in the newer server to create a direct comparison to the older model, the company said at a briefing on Tuesday. Nvidia has been working to speed up the connections of chips inside its servers, which is important in AI work where a chatbot runs on multiple chips at once. The second benchmark is also based on an open-source AI model built by Meta and the test aims to more closely simulate the performance expectations set by consumer AI applications such as ChatGPT. The goal is to tighten the response time for the benchmark and make it close to an instant response.
Yahoo
28-03-2025
- Health
- Yahoo
New study uncovers alarming effects of AI systems on human health: 'It's a public health issue we need to address urgently'
Artificial intelligence is transforming industries, but a hidden cost is emerging: its pollution footprint could put thousands of lives at risk. A new study warns that AI-driven data centers generate significant air pollution, contributing to long-term health issues across the United States. Researchers from the University of California, Riverside and California Institute of Technology have uncovered that AI-driven data centers are fueling a surge in air pollution with major health implications. As AI demand skyrockets, these massive facilities require enormous amounts of electricity, much of it generated by dirty fuel-burning power plants and diesel backup generators. By 2030, pollution from these power sources could cause up to 1,300 premature deaths per year in the U.S., according to the study. The public health costs associated with this pollution could reach nearly $20 billion annually, with increased risks of cancer, asthma, and respiratory diseases in affected communities. "If you have family members with asthma or other health conditions, the air pollution from these data centers could be affecting them right now. It's a public health issue we need to address urgently," said UC Riverside associate professor Shaolei Ren, a corresponding author of the study. Yet despite these findings, many major tech companies do not account for air pollution data in their sustainability reports, focusing instead on carbon pollution and water usage. The harms of AI-driven pollution aren't just local — they extend far beyond the communities where data centers are built. Backup generators in Northern Virginia, for example, have been linked to air pollution that spreads across Maryland, Pennsylvania, and New Jersey, adding up to $260 million in regional health costs. If pollution reaches its permitted maximum, that figure could soar to $2.6 billion per year. Do you worry about air pollution in your town? All the time Often Only sometimes Never Click your choice to see results and speak your mind. Additionally, the study found that training AI models, such as Meta's Llama 3.1, can generate as much pollution as 10,000 round trips between Los Angeles and New York by car. As AI technology continues to expand, its pollution footprint could rival that of the U.S. steel industry and even surpass pollution from all vehicles in California. Despite these troubling statistics, solutions exist to curb AI's air pollution footprint. Stronger regulations could require tech companies to transition away from dirty fuel-based power sources and toward clean energy alternatives such as wind and solar. Policies that incentivize renewable energy adoption, such as tax breaks for clean energy-powered data centers, could make a major difference. Scientific innovation is also opening new doors. Researchers are developing methods to convert air pollution into usable fuel, turning harmful pollution into alternative energy sources. Join our free newsletter for weekly updates on the latest innovations improving our lives and shaping our future, and don't miss this cool list of easy ways to help yourself while helping the planet.
Zawya
21-03-2025
- Business
- Zawya
Trend Micro to Open-source Model and AI Agent to Drive the Future of Agentic Cybersecurity
Leveraging NVIDIA AI to deliver powerful proactive security and scalable threat prevention for GenAI applications Enhance security posture: enabling teams to precisely anticipate risk across the entire attack surface. enabling teams to precisely anticipate risk across the entire attack surface. Reduce alert overload : alleviate fatigue for SecOps teams through more accurate prioritization. : alleviate fatigue for SecOps teams through more accurate prioritization. Save developers' time : overcome security skills shortages by providing actionable insights to help them identify and remediate risks. : overcome security skills shortages by providing actionable insights to help them identify and remediate risks. Unleash greater value: deliver more powerful insights from existing risk sensors. HONG KONG SAR - Media OutReach Newswire - 21 March 2025 - Trend Micro Incorporated TSE: 4704 ), a global cybersecurity leader, today announced the open-sourcing of Trend Cybertron, an AI model and agent framework designed to accelerate the development of autonomous cybersecurity agents. As one of the first specialized cybersecurity LLMs, it provides organizations and researchers worldwide access to advanced cybersecurity capabilities at no cost. The specialized Trend Cybertron model is fine-tuned using Llama 3.1 and supports rapid, reliable deployment with NVIDIA NIM inference microservices on NVIDIA accelerated infrastructure."The secret sauce of Trend Cybertron is the data it continuously learns from, fine-tuned for optimized threat detection and mitigation. By bringing to bear the very highest quality threat data and NVIDIA's industry-leading AI expertise, we've madeproactive security a reality, enabling us to predict and prevent threats like never before. This innovation isn't just a win for our customers—it's about making the entire digital, connected world a safer place."Global organizations are struggling to innovate and grow while being weighed down by security challenges, fragmented point solutions, and an overwhelming flood of threat alerts. The complexity demands a shift to a proactive approach. Powered by NVIDIA AI at the core, Trend Cybertron moves beyond chasing threats, applying intelligent, decision-making AI agents to predict and respond.: "With the ability to understand, reason and take action, AI agents give organizations a powerful new cybersecurity tool. Agentic AI security agents built with the Trend Cybertron model and framework using NVIDIA AI can analyze massive amounts of data in real time to detect potential threats, adapt dynamically, and respond autonomously. "To operationalize this vision, Trend has built its agentic AI strategy with NVIDIA AI software to accelerate cybersecurity automation, helping organizations proactively manage risks with resource scanning, risk assessment, priority-based reasoning, and actionable remediation can apply multiple blueprints and seamlessly integrate AI agents to automate security tasks with resource scanning, proactively mitigate threats, and scale their defenses to manage threats more effectively, all within the NVIDIA ecosystem. Specifically, this will help:Trend Cybertron is designed to proactively manage risk, leveraging threat intelligence from over 250 million sensors worldwide—the broadest in the industry. It interprets user queries, generates actionable plans, and performs a holistic risk assessment by retrieving real-time cybersecurity intelligence from Trend Micro's cloud, ultimately providing with tailored recommendations and best practices to secure an enterprise's AI provided support and AI microservices for developing and deploying the model. Trend Micro trained and optimized the Trend Cybertron model for inference using NVIDIA DGX supercomputing to reduce the time required to fine-tune the Cybertron currently consists of an 8-billion-parameter AI model and an initial specialized AI agent, with additional models and agents in development to expand its cybersecurity capabilities. A larger and more advanced version with 70-billion-parameter of the model is planned for the near future to address the cybersecurity challenges of #trendmicro #trendvisionone #visionone #cybersecurity #trendcybertron #cybertron The issuer is solely responsible for the content of this announcement. Trend Micro Trend Micro, a global cybersecurity leader, helps make the world safe for exchanging digital information. Fueled by decades of security expertise, global threat research, and continuous innovation, Trend Micro's cybersecurity platform protects hundreds of thousands of organizations and millions of individuals across clouds, networks, devices, and endpoints. As a leader in cloud and enterprise cybersecurity, the platform delivers a powerful range of advanced threat defense techniques optimized for environments like AWS, Microsoft, and Google, and central visibility for better, faster detection and response. With 7,000 employees across 65 countries, Trend Micro enables organizations to simplify and secure their connected world. Trend Micro
