Jun 29, 2024

• Microsoft says that it's okay to steal web content it because it's 'freeware.'

News Highlights

• Microsoft's AI CEO, Mustafa Suleyman, claims content on the web is 'freeware' for AI training unless explicitly stated otherwise, sparking debate on the ethics and legality of using web content to train AI models.

• Open-LLM performances have hit a plateau, indicating a slowdown in the pace of advancements in large language models as observed by the Hugging Face Space's latest analysis.

• EvTexture enhances video super-resolution by integrating event-driven texture information, achieving significant improvements on the Vid4 and REDS4 test sets with a $4\times$ upsampling rate. Paper

• GEMM tuning on AMD MI300x GPUs significantly enhances throughput and latency for AI models, with throughput improvements up to 7.2x for models like LLaMA-2-70B.

• llama-agents is an async-first framework designed for building and deploying multi-agent systems, facilitating features like multi-agent communication and human-in-the-loop processes.

• Microsoft AI's CEO, Mustafa Suleyman, declared online content as 'freeware' for training AI models, sparking legal challenges from content creators and organizations like the Center for Investigative Reporting and the New York Times.

• Finetuning Large Language Models (LLMs) on a synthetic dataset significantly enhances their ability to retrieve information and reason over long-context inputs, as demonstrated in experiments with GPT-3.5 Turbo and Mistral 7B.

• Microsoft has unveiled the Skeleton Key attack, a method that can bypass AI model safety measures to generate harmful content, such as instructions for making a Molotov cocktail.

• Context-augmented Retrieval (CAR) enhances Large Language Model (LLM) performance by partitioning the vector database based on real-time classification, ensuring quick and relevant information retrieval.

• The Segment any Text (SaT) model introduces a new pretraining scheme to enhance robustness by reducing reliance on punctuation, addressing a common shortfall in existing sentence segmentation methods.

• The Vision Language Group at IIT Roorkee has crafted detailed summaries of deep learning papers from 2016 to 2024, covering major conferences like NeurIPS, CVPR, ICCV, and ICML.