This week brought a whirlwind of developments in the world of AI, from OpenAI’s record-breaking funding round and Nvidia’s foray into open-source LLMs to Google’s multilingual Gemini Live and unsettling privacy concerns raised by Meta Smart Glasses. Read on to get the scoop on these major announcements.
Phison’s aiDAPTIV+ is a groundbreaking hybrid software and hardware solution that enhances LLM training by utilizing SSD storage to expand GPU memory capacity. This innovative approach allows for massive model support with low latency, enabling workstations and servers to handle workloads previously restricted to data centers. aiDAPTIV+ empowers users to train models like Llama-3 70B and Falcon 180B with cost-effective scalability and ease of integration.
The Kerala Law Entrance Exam (KLEE) 2024 for LLM registration closes today, August 6. Eligible candidates can apply online at cee.kerala.gov.in. The exam is scheduled for August 17, with admit cards releasing on August 12. Find out the eligibility criteria, application process, and exam pattern.
Meta’s Llama 3 405B model training took 54 days on a cluster of 16,384 NVIDIA H100 AI GPUs, facing numerous hardware challenges. The team overcame these hurdles with in-house tools and efficient strategies, achieving a 90% effective training time.
This article compares two leading large language models (LLMs): Anthropic’s Claude 3.5 Sonnet and OpenAI’s GPT-4o. It explores their strengths and weaknesses across various tasks, including accuracy, response speed, pricing, and capabilities. The article aims to provide a comprehensive overview to help readers choose the LLM best suited for their needs and budget.
Researchers at the University of California, Santa Cruz have found a way to significantly reduce the energy consumption of large language models (LLMs) without sacrificing performance. By simplifying the neural network’s operations and utilizing custom hardware, they have achieved a 50x improvement in energy efficiency, reducing power consumption from 700W to just 13W.
