Microsoft Azure AI Introduces GPT-4 and New Virtual Machines, Enhancing AI Capabilities

2023-08-09

Microsoft and OpenAI have been collaborating for a long time, from Microsoft's initial investment in the company behind ChatGPT to embedding GPT services into the Azure AI platform. In a blog post detailing updates to the Azure AI platform on Monday, Microsoft stated that Azure OpenAI services are at the "forefront" of the generative AI revolution, while Azure AI infrastructure is the "backbone". OpenAI's two major changes and upgrades to virtual machine hardware demonstrate Microsoft's commitment to expanding generative AI in more areas. Azure OpenAI has launched new models, including GPT-4. GPT-4 and GPT-35-Turbo from OpenAI are now available through Azure OpenAI in four new regions, including Canada East, Japan East, UK South, and another region in the US East (East US 2 on the availability map). GPT-4 is currently the most advanced generative AI model offered by OpenAI. Azure OpenAI services are used by approximately 11,000 customers for tasks such as customer service, content writing, and document analysis. "We see that Azure OpenAI's ChatGPT editor is helping users create more relevant, personalized, and even more creative content," said Kevin Souers, Chief Product Officer at Aprimo. Existing Azure OpenAI customers can now join the waitlist to access GPT-4. What does the new NVIDIA virtual machine mean for Azure customers? A new virtual machine series, Azure ND H100 v5, is now generally available to existing enterprise customers in the US East and US Central South Azure regions. These virtual machines are designed to help organizations design and run generative AI applications. This new hardware includes the NVIDIA H100 Tensor Core GPU and NVIDIA Quantum-2 InfiniBand network, optimized for AI performance. The low-latency network includes 400Gb/s NVIDIA Quantum-2 ConnectX-7 InfiniBand per GPU, achieving a bandwidth of 3.2Tb/s per node, delivering supercomputer-level performance per virtual machine. The PCIe Gen5 data transfer standard in the ND H100 v5 virtual machine boosts GPU performance to a bandwidth of 64GB/s per GPU, improving performance between the CPU and GPU. Microsoft states that operations on the ND H100 v5 virtual machine will be faster, and some large language models will see a doubling of speed when running on it. Red team testing needs to adapt to generative AI behavior. Security remains a concern when it comes to AI. Microsoft assures customers, "We combine powerful security systems with human feedback mechanisms to responsibly handle harmful input." Microsoft also encourages red team testing of AI applications, inviting ethical hackers to play the role of threat actors to assess the attack risks AI applications may face. Microsoft's guidelines for red teams preparing to work with AI include: - Focus on security and responsible AI outcomes, ensuring output results are not malicious or dangerous. - Consider that not only malicious interactions but also well-intentioned interactions may lead to unwanted output results. - Ensure both defense and attack techniques are comprehensive. Competitors of Azure AI include Amazon Web Services, IBM Watson, Google AI, DataRobot's custom AI model services, Salesforce's Einstein AI for marketing, ServiceNow AIOps for IT operations management, Oracle Cloud Infrastructure, and H2O.ai.

RECENT AI NEWS

RECENT AI TOOLS