News from the AI & ML world
DeeperML - #customers
|
@cloud.google.com
//
Google Cloud is enhancing its AI Hypercomputer to accelerate AI inference workloads, focusing on maximizing performance and reducing costs for generative AI applications. At Google Cloud Next 25, updates to AI Hypercomputer's inference capabilities were shared, showcasing Google's newest Tensor Processing Unit (TPU) called Ironwood, designed for inference. Software enhancements include simple and performant inference using vLLM on TPU and the latest GKE inference capabilities such as GKE Inference Gateway and GKE Inference Quickstart. Google is paving the way for the next phase of AI's rapid evolution with the AI Hypercomputer.
Google's JetStream inference engine incorporates new performance optimizations, integrating Pathways for ultra-low latency multi-host, disaggregated serving. The sixth-generation Trillium TPU exceeds throughput performance by 2.9x for Llama 2 70B and 2.8x for Mixtral 8x7B compared to TPU v5e. Google’s JAX inference engine maximizes performance and reduces inference costs by offering more choice when serving LLMs on TPU. JetStream throughput is improved, achieving 1703 token/s on Llama 3.1 405B on Trillium. Google is also intensifying its efforts to combat online scams by integrating artificial intelligence across Search, Chrome, and Android. AI is central to Google's anti-scam strategy, blocking hundreds of millions of scam results daily and identifying more fraudulent pages. Gemini Nano provides instant detection of high-risk websites, helping counter new and evolving scams across platforms. Google has long used AI to detect and block scams, including fake tech support, fraudulent financial services, and phishing links. Recent updates to AI classifiers now allow the company to detect 20 times more scam pages, improving the quality of search results by reducing exposure to harmful sites. Recommended read:
References :
@cloud.google.com
//
Google is significantly expanding the AI and ML capabilities within its BigQuery and Vertex AI platforms. BigQuery is receiving a boost with the integration of the TimesFM forecasting model, a state-of-the-art, pre-trained model from Google Research designed to simplify forecasting problems. This managed and scalable engine enables users to generate forecasts for both single and millions of time series within a single query. Additionally, BigQuery now supports structured data extraction and generation using large language models (LLMs) through the AI.GENERATE_TABLE function, alongside new row-wise inference functions, expanded model choice with Gemini and OSS models, and the general availability of the Contribution Analysis feature.
NotebookLM is also seeing expansion with the "Audio Overviews" feature now available in approximately 75 languages. This feature, powered by Gemini, allows users to listen to AI-generated summaries of documents, slides, web pages, and YouTube transcripts in multiple languages. This feature distills any mix of documents into a scripted back-and-forth between two synthetic hosts. Users can direct tone and depth through a prompt and then download an MP3 or keep playback inside the notebook. Early testers report that multilingual voices make long reading lists easier to digest on commutes and provide an alternative channel for blind or low-vision audiences. Furthermore, Google is experimenting with AI-powered language learning formats through its “Little Language Lessons,” integrated directly into NotebookLM and running on Gemini. These tools support situational learning, generating content dynamically based on user-described scenarios, rather than relying on fixed vocabulary lists. Google is also preparing new Gemini AI subscription tiers, potentially including a "Gemini Ultra" plan, evidenced by code discoveries in the Gemini web interface referencing distinct tiers with varying capabilities and usage limits. Recommended read:
References :
Kailyn Sylvester@Microsoft Security Blog
//
Microsoft is actively enhancing its AI integration across Azure and Copilot, introducing new features and programs to support both enterprise security and AI innovation. The company is now offering the "Llama 4 herd" within Azure AI Foundry and Azure Databricks, providing users with more AI tools and resources. Simultaneously, Microsoft is working to improve Copilot's capabilities by integrating enhanced security measures and features designed to facilitate AI adoption within organizations. These advancements reflect Microsoft's commitment to making AI more accessible and secure for its users.
Microsoft is also working with external partners to foster AI development. Microsoft Hong Kong has collaborated with Cyberport to launch the "Cyberport x Microsoft AI Partnership Programme," aimed at nurturing local start-ups. This program provides benefits like solution support, expert guidance, and business matching opportunities to promising Hong Kong-based companies. Six companies were selected to participate, showcasing innovative AI solutions in healthcare, insurance, risk management, and corporate sustainability. In addition to external partnerships, Microsoft is focused on internal security enhancements related to AI. Microsoft Copilot utilizes classification labels as part of Microsoft Information Protection (MIP) to safeguard sensitive information, ensuring data security and regulatory compliance. These labels, applied manually, automatically, or suggested by Copilot, categorize data based on sensitivity levels, such as public, internal, or confidential. Furthermore, Microsoft is hosting a "Tech Accelerator: Azure Security and AI Adoption" event on April 22, 2025, designed to equip developers and cloud architects with the essential guidance and resources needed to securely plan, build, manage, and optimize their Azure deployments and AI projects. Recommended read:
References :
Aminu Abdullahi@eWEEK
//
References:
eWEEK
, www.techradar.com
,
Yum! Brands, the parent company of fast-food giants Taco Bell, KFC, and Pizza Hut, is partnering with Nvidia to integrate AI-powered order-taking systems at 500 locations by the end of 2025. The aim is to enhance voice recording, improve drive-thru efficiency, and streamline overall restaurant operations. This initiative marks a significant step in the company’s push to digitize its operations and stay ahead in the competitive quick-service restaurant industry.
These AI voice agents, built using Nvidia’s AI tools, can understand natural speech, process complex orders, and even suggest add-ons like extra fries or a dessert. Yum! is also using Nvidia’s computer vision tech to help restaurants analyze drive-thru traffic and improve speed during peak hours, potentially reducing wait times and improving staffing decisions. The technology could also determine whether the food being served matches what was ordered by analyzing images from existing CCTV cameras. Recommended read:
References :
|
BenchmarksBlogsResearch Tools |