LLM Pulse - Dec 2, 2024

New Releases & Updates

NVIDIA Introduces Hymba 1.5B: A Hybrid Small Language Model Outperforming Llama 3.2 and SmolLM v2: Large language models (LLMs) like GPT-4 and Llama-2 are powerful but require significant computational resources, making them impractical for smaller devices. Attention-based transformer models, in particular, have high memory demands and quadratic computational complexity, which limits their efficiency. Read More

Apple Prepares to Revolutionize Siri with AI-Powered “LLM Siri” by 2026: Before being acquired by Apple in 2011, Siri was a standalone app designed to revolutionize the way users interacted with their iOS devices. Under Apple’s leadership, Siri became a trailblazer in voice assistant technology. Read More

S&P Global Launches Kensho  LLM -ready API beta, Making its Structured Data Accessible: for Generative AI S&P Global announced the launch of a new solution in open beta that enables customers to access several high-priority S&P Global datasets for generative AI (GenAI) use cases. Kensho LLM-ready API integrates seamlessly with large language models (LLMs) like GPT, Gemini, or Claude, allowing customers to use natural language to query S&P Global's tabular datasets. Read More

SoftBank Corp. - Building a Japanese-based  LLM  for the Next Leap: SoftBank is advancing its technology strategy to realize "Next-generation Social Infrastructure" to support a "society that coexists with AI" and enhance corporate value. As part of this initiative, in August 2023 it launched SB Intuitions Corp., a wholly owned subsidiary to develop homegrown Large Language Models (LLM) specialized for the Japanese language, as well as develop, market and provide generative AI services. Read More

NVIDIA Megatron-LM Powers 172 Billion Parameter LLM for Japanese Language Proficiency: NVIDIA's Megatron-LM is at the forefront of a significant development in the field of Natural Language Processing (NLP) by powering a large language model (LLM) with 172 billion parameters, aimed at enhancing Japanese language processing capabilities. This initiative is part of the Generative AI Accelerator Challenge (GENIAC) project, as reported by NVIDIA. Read More

SK Telecom to Introduce AI-Powered Customer Service Utilizing Proprietary Telco LLM and LMM: SK Telecom launches AI Customer Service Support System, utilizing its proprietary Telco large language model (LLM) and large multimodal model (LMM). The system will be gradually deployed in customer consultations to enhance efficiency and improve customer service. AI enhances service efficiency by providing AI Knowledge Search Assistant, Intelligent Document Processing, and Automated Post-Processing System for Consultation Results. Read More

NEC develops Agentic AI to boost productivity through automation of advanced specialized tasks: NEC said it will offer AI agents that autonomously execute tasks by linking various AIs, including generative AI, and IT services from January 2025. When a user inputs the task they wish to request, NEC's AI agent autonomously breaks down the tasks and designs the necessary business processes. It then selects the most suitable AI and IT services for each task and automatically executes the task. Read More

Taishin Bank develops own LLM with OneDegree’s help: What is zero trust in cybersecurity? The bank has partnered with OneDegree Global to provide the cybersecurity. Read More

Huawei Cloud Empowers DTGO to Launch the World's First Large Language Model: Huawei Cloud and DTGO Corporation Limited (DTGO), a prominent business group dedicated to social and environmental impact, have announced the successful launch of the world's first large language model (LLM) capable of efficient performance. Read More

Alibaba Marco-o1: Advancing LLM reasoning capabilities: Alibaba has announced Marco-o1, a large language model (LLM) designed to tackle both conventional and open-ended problem-solving tasks.Marco-o1, from Alibaba’s MarcoPolo team, represents another step forward in the ability of AI to handle complex reasoning challenges—particularly in maths, physics, coding, and areas where clear standards may be absent. Read More

xAI’s standalone Grok app is coming soon: Elon Musk’s xAI may be a newcomer to the artificial intelligence segment, but its influence has grown significantly in the previous months. A huge part of xAI’s growing prominence is due to Grok, since the large language model has real-time access to data on X posts. Grok, however, is fully tied to X’s ecosystem for now. Read More 

Research and Technology

Multilingual and open source: OpenGPT-X research project releases large language model: The large language model of the OpenGPT-X research project is now available for download on Hugging Face: "Teuken-7B" has been trained from scratch in all 24 official languages of the European Union and contains 7 billion parameters. Read More

NVIDIA's TensorRT- LLM  Multiblock Attention Enhances AI Inference on HGX H200: In a significant development for AI inference, NVIDIA has unveiled its TensorRT-LLM multiblock attention feature, which substantially enhances throughput on the NVIDIA HGX H200 platform. According to NVIDIA, this innovation boosts throughput by more than 3x for long sequence lengths, addressing the increasing demands of modern generative AI models. Read More

Need a Trillion-Parameter LLM? Google Cloud Is for You: At KubeCon+CloudNativeCon North America earlier this month, Google Cloud announced it had upgraded its Google Kubernetes Engine (GKE) to support clusters of up to 65,000 nodes. That’s a big leap up from its previous limit of 15,000 nodes. Read More

Study reveals cost-efficiency strategy for LLM deployment: Researchers from the Icahn School of Medicine at Mount Sinai have identified avenues for cost-effective large language model deployment at health system scale, according to a recent Npj Digital Medicine study. Read More

A cost-effective LLM use in hospitals: Study Large language models can complete 50 simultaneous tasks and drive a seventeen fold cost reduction, but any additional tasks will cause performance deterioration, according to a study published Nov. 18 in Nature. Read More

DXC transforms data exploration for their oil and gas customers with LLM-powered tools: One of the sectors DXC has deep expertise in is energy. The oil and gas industry relies on discovering new drilling sites to drive growth. Data-driven insights can accelerate the process of identifying potential locations. Read More

NVIDIA Fugatto - The Future Of Sound Design? : LLM Focussing on speech, singing, and sound effects: As more and more LLM (Large Language Models) that power experiments with AI continue to appear, NVIDIA (whose GPUs generally power this research) has unveiled Fugatto, a generative AI model Read More

Salesforce CEO says  LLM  ‘upper limits’ reached, future of AI is agents: Salesforce CEO Marc Benioff is the latest to tout AI agents as the next advancement of artificial intelligence technologies. Marc Benioff, CEO of American cloud computing software firm Salesforce, said the future of artificial intelligence lies in autonomous agents rather than large language models (LLMs) in the form of AI chatbots. Read More

Comparing Llama and GPT: Open-Source Versus Closed-Source AI Development: As it stands, GPT-4 is the king of general-purpose large language models. But for building specialized LLM-based products, Llama 2 might prove superior due to its comparable or superior factual accuracy. Read More

LLMPhy Revolutionizes Physical Reasoning by Combining AI and Physics Simulation: LLMPhy redefines problem-solving, blending advanced AI reasoning with physics simulation to tackle real-world challenges in object dynamics and stability. In an article recently submitted to the arXiv preprint* server, researchers at Mitsubishi Electric Research Labs (MERL) introduced a zero-shot optimization framework for combining large language models (LLM) for physical reasoning, LLMPhy  Read More.

Knostic research unveils timing-based vulnerabilities in AI large language models: New research out today from Knostic Inc., a startup that provides need-to-know-based access controls for large language models, details a new category of vulnerabilities in LLMs that can be used by attackers to bypass guardrails and extract sensitive information. Read More.

Other News 

Large language models not fit for real-world use, scientists warn — even slight changes cause their world models to collapse: Large language model AIs might seem smart on a surface level but they struggle to actually understand the real world and model it accurately, a new study finds. Read More

The Challenges and Opportunities in Developing LLM Solutions: The advent of Large Language Models (LLMs) marks a profound leap in computational intelligence, akin to the foundational breakthroughs in early computing. Read More

AI might not dream, but it does hallucinate: What wrong answers can tell you about choosing the right LLM: As AI becomes an ever-present part of our daily lives, questioning and scepticism of its ability – and more recently, whether it all exists in a bubble – has become commonplace. But perhaps the most pressing issue is the phenomenon of AI “hallucinations.” Read More

Does Mistral have what it takes to win the LLM market? Mistral’s team of just over 100 people started monetising its products at the beginning of 2024; the company is on track to finish the year with €30m annual recurring revenue (ARR), two sources with direct knowledge of the matter, who wished to remain anonymous to protect relationships, tell Sifted. Mistral declined to comment. Read More

Advantech forges strategic partnership with Namla to scale the deployment of AI & LLM  at the Edge: Advantech, a leading industrial edge AI platform provider, is excited to announce its partnership with Namla, a provider of a Cloud Native Edge Orchestration platform. This collaboration represents a significant milestone in the Edge AI industry, optimizing Edge infrastructure deployment and enabling the full potential of Cloud Native architecture to be harnessed on robust Edge AI hardware. Read More

Beyond the LLM Plateau: How Specialized Language Models Advance AI: Given the huge amounts of venture capital going into LLMs and chips at the tech cartel and, the internal investments at the big consultants and corporations, what’s left of actual free market capitalism needs to spend its money wisely. Read More

Securing AI: What the OWASP LLM Top 10 Gets Right – and What It Misses: Securing AI systems is a pressing concern for CIOs and CISOs due to AI and LLMs’ increasingly vital role in businesses. Thus, they instinctively turn to Open Web Application Security Project (OWASP) for guidance. Read More