This month, the Azure AI portfolio is unveiling enhanced capabilities for making more informed decisions and adapting to build and scale AI solutions?
Over 60,000 prospects together with AT&T, H&R Block, Volvo, Grammarly, Harvey, Leya, and extra leverage Microsoft Azure AI to drive AI transformation. The growing enthusiasm for AI’s widespread integration across industries is palpable. What follows is an overview of innovative features that enable the creation and scaling of more advanced and adaptable AI solutions. Key updates embrace:
Azure OpenAI Knowledge Zones: A New Era in AI-Powered Insights
Azure OpenAI Knowledge Zones have revolutionized the way organizations access and utilize AI-driven knowledge. With a strong presence in both the United States and the European Union, these zones are poised to transform industries and ecosystems alike.
By providing seamless connections to Azure’s vast array of AI services, the Knowledge Zones empower developers to build innovative applications that drive business value. From natural language processing to computer vision, organizations can leverage the power of AI to gain critical insights and make data-driven decisions.
In the United States, the Knowledge Zones are strategically located in major hubs such as New York City, San Francisco, and Seattle, allowing for easy access to top-tier talent, cutting-edge research institutions, and a vast network of potential partners and clients.
Similarly, in the European Union, the Knowledge Zones are situated in key cities like London, Berlin, and Paris, placing them at the heart of Europe’s thriving tech ecosystem. This proximity enables seamless collaboration with local innovators, researchers, and industry leaders, fostering a culture of innovation and entrepreneurship.
As AI continues to evolve and become an increasingly integral part of our daily lives, the Azure OpenAI Knowledge Zones will remain at the forefront, driving progress, and shaping the future of industries across both continents.
We are delighted to introduce a groundbreaking deployment option that empowers businesses to exercise greater control and oversight over their sensitive data’s privacy and residency requirements, ensuring enhanced security and compliance. Geared towards organizations operating within the US and EU, Knowledge Zones enable prospects to process and store data within specific geographic parameters, thereby ensuring compliance with regional data residency requirements while preserving optimal performance. With a presence across multiple regions, Knowledge Zones strike a balance between the economies of scale from global deployments and the tailored management of regional implementations, enabling enterprises to effectively manage their AI capabilities while maintaining speed and security.
The new function simplifies complex information management by offering a streamlined solution that enables faster deployment of AI innovations from Azure OpenAI Service, allowing for greater processing capacity and earlier access to the latest AI models. Enterprises can now leverage Azure’s robust infrastructure to securely scale their AI capabilities and meet strict information residency requirements. Knowledge Zones is now available for Customary (Pay-As-You-Go) customers, with a forthcoming rollout for Provisioned users.
Azure OpenAI Service updates
This month, we unveiled the Azure OpenAI Batch API, empowering builders to process large-scale and high-volume tasks more efficiently through dedicated quotas, a 24-hour turnaround, and a 50% reduced pricing compared to our standard international rates. Within McKesson, Ontada has successfully harnessed the power of Batch API to process immense volumes of patient data across oncology facilities in the United States with remarkable efficiency and cost-effectiveness.
Sagaran Moodley, Chief Innovation and Know-how Officer at Ontada.
Additionally, we’ve enabled immediate caching in our Azure OpenAI Service. This feature optimizes prices and latency by reusing recently seen input tokens with our builders. This function is particularly useful for situations where the same contextual information is used repeatedly, such as coding or conversing with AI assistants. Immediate caching enables swift and efficient rendering of customary pages.
We’re simplifying international deployment provisioning by reducing the initial allocation for GPT-4 fashion models to 15 Provisioned Throughput Units (PTUs), with subsequent increases of 5 PTUs. To further expand accessibility to the Azure OpenAI Service, we’re cutting the price of Provisioned International Hourly by half. Pricing strategies for artificial intelligence (AI) deployments must strike a delicate balance between revenue goals and customer expectations. To achieve this equilibrium, organizations should consider the total cost of ownership (TCO), including infrastructure expenses, training data acquisition costs, and maintenance fees. A tiered pricing structure based on complexity, scalability, and customization requirements can also help ensure a fair return on investment for all parties involved. Moreover, AI-driven services offering variable pricing models that adapt to usage patterns or time-of-day fluctuations can provide an added layer of flexibility. Ultimately, successful price management hinges on fostering open communication with customers and being prepared to adjust the pricing strategy as market conditions evolve?
We’re introducing a 99% latency service level agreement (SLA) for our token era. This latency service level agreement guarantees token generation at consistent and rapid speeds, especially during high-volume scenarios.
New fashions and customization
We expand our mannequin assortment by incorporating cutting-edge fashion trends into our catalog. New fashion collections are available this month, featuring designs from esteemed brands Mistral and Cohere, among others. We’re also introducing customization options for the Phi 3.5 family of designs.
- Comprising advanced multimodal medical imaging techniques, complemented by MedImageInsight for image analysis, MedImageParse for modality-agnostic image segmentation, and CXRReportGen that enables the generation of comprehensive, structured reports. In partnership with Microsoft Analytics and industry leaders, these models have been engineered to be easily fine-tuned and tailored by healthcare organizations, streamlining the process of building bespoke solutions that cater to unique needs while minimizing computational and data requirements. What discoveries are being made at this present moment?
- The Ministral 3B marks a significant milestone within the sub-10 billion cluster, excelling in the realms of information processing, common sense reasoning, function invocation, and efficiency. With support for up to 128k context size, these styles are tailored to accommodate a diverse range of applications – from streamlining agent-based workflows to building specialized task forces. Used in conjunction with larger linguistic frameworks such as Mistral Massive, Ministral 3B can effectively serve as a modular mediator for function invocation within complex, multi-step workflow architectures.
- Now available in the Azure AI Model Catalog, Embed 3, Cohere’s cutting-edge AI search model, has been upgraded to support multimodal capabilities. Embedding innovative technology that effortlessly transforms diverse data formats into valuable insights, Embed 3 empowers businesses to unlock the full potential of their vast datasets, facilitating seamless search and analysis capabilities.
The Embed 3 model stands out as the most effective and successful multimodal embedding approach available, revolutionizing how businesses search complex assets like stories, product catalogs, and design files.
- With Phi-3.5-mini and Phi-3.5-MoE alongside. Philosophical household fashions are well-suited for customization, enabling individuals to tailor their approach to optimize performance across various scenarios, such as learning a new skill or job, or improving consistency and quality of response. With their minuscule computational requirements, coupled with seamless integration into cloud and edge environments, Phi-3.5 variants offer a cost-effective and eco-friendly alternative to larger models of the same size or next in line. Adoption of the Phi-3.5 household concept is accelerating, with applications emerging that combine edge reasoning and scenario-based analysis for both connected and disconnected scenarios. Currently, builders can refine Phi-3.5-mini and Phi-3.5-MoE models on the fly using a serverless endpoint provided by Mannequin, enabling seamless experimentation and iteration.
AI app improvement
We’re designing Azure AI as a flexible, modular architecture, enabling developers to seamlessly transition from idea to implementation to deployment in the cloud. Builders can now seamlessly discover and deploy Azure AI models directly from GitHub Marketplace through the Azure AI model inference API. Developers can experiment with diverse styles and assess model performance freely in a playground setting, then seamlessly deploy and customize to their Azure account, scaling from token-based usage to paid endpoints with enhanced security and monitoring capabilities while maintaining code integrity.
We have also accelerated the development of our AI application by introducing new enhancements. Developers can leverage these versatile templates across various integrated development environments (IDEs), including GitHub Codespaces, Visual Studio Code, and Microsoft Visual Studio. Templates offer versatility through a wide range of styles, frameworks, languages, and supplier-provided options from respected brands such as Arize, LangChain, LlamaIndex, and Pinecone. Builders can deploy full applications or start with individual components, provision services across Azure and partner organizations.
What drives us is empowering every builder worldwide to build smarter with AI’s limitless potential? With these updates, builders can swiftly kick off in their preferred environment, choose the deployment option that best aligns with their needs and scale AI capabilities with confidence.
Building on advances in cloud-based infrastructure and containerization, innovative approaches are emerging to construct robust, scalable, and secure AI applications that cater to the demands of modern enterprises.
At Microsoft, our focus is on empowering users to leverage and build AI that is secure, trustworthy, and tailored to their needs. Currently, we’re thrilled to announce the introduction of two innovative features that empower you to build and grow AI solutions with confidence.
The Azure AI model catalog offers more than 1,700 pre-trained models for developers to explore, evaluate, customize, and implement seamlessly. While a vast array of choices fosters creativity and flexibility, it may also pose significant obstacles for organizations seeking to ensure that all deployed models conform to their internal policies, security protocols, and regulatory demands. Azure AI directors now have the ability to directly access the Azure AI model catalog, streamlining the process of selecting and governing models for their organization. The revised text is: This package includes pre-built insurance policies for Fashion-as-a-Service (MaaS) and Fashion-as-a-Platform (MaAP) deployments, while detailed information enables the creation of tailored insurance policies for Azure OpenAI Service and other AI providers. Collected together, these insurance policies collectively provide comprehensive coverage for establishing a standardized model record and its seamless integration across platforms.
To customize fashion and function, developers may access resources both on-premises and those not supported by private endpoints but still situated within their custom Azure virtual network (VNET). A load balancer that primarily routes requests based on the URL of HTTPS traffic. The Utility Gateway enables secure communication between a private network and external sources by establishing a tunnel for HTTP and HTTPS protocols. Currently, it is validated to establish a secure connection to JFrog Artifactory, Snowflake Database, and Private APIs. Now generally available in public preview, developers can leverage on-premises or custom Virtual Network (VNET) sources for their training, fine-tuning, and inferencing scenarios without compromising their security posture.
Azure AI, harnessing the power of machine learning and artificial intelligence, empowers organizations to drive innovation, streamline operations, and make data-driven decisions. With a comprehensive suite of AI tools and services, you can build, deploy, and manage intelligent applications that transform your business. From natural language processing to computer vision, predictive analytics to recommendation systems, Azure AI helps you unlock the potential of your data and stay ahead of the competition.
Over the past half year at Azure AI, I’ve had the privilege of driving cutting-edge AI advancements, witnessing developers build game-changing experiences leveraging our tools, and gaining valuable insights from customers and partners. What lies ahead excites me immensely. Join us as we delve into the latest updates from ?
