Microsoft is implementing a series of upgrades to its AI stack, including Azure AI Studio and Copilot Studio, developer platforms for building apps that leverage generative AI. At its Build 2024 conference, the company also announced it is going into preview with its Azure virtual machines (VMs), the first platform to use the powerful Arm-based Cobalt 100 cloud processors developed by Microsoft, which shared that plan at its November Ignite conference. Other Build highlights include the general availability of OpenAI’s new GPT-4o flagship model across the Azure OpenAI Service and the new multimodal Phi-3-vision SLM.

At Microsoft Build 2024 the company announced the general availability of Azure AI Studio and said it will soon add “model-as-a-service” (MaaS) capability, allowing developers to “build generative AI-powered apps using pay-as-you-go inference APIs — the APIs through which developers can access and fine-tune generative AI models hosted on Azure infrastructure,” TechCrunch writes.

MaaS now in preview include G42’s Core42 JAIS series, a leader in Arabic-language AI that also works in English. Also, the predictive TimeGen-1 from Nixtla. Models from AI21 Labs, Bria AI, Gretel Labs, NTT DATA, Stability AI as well as the Rerank models from Cohere, Microsoft explains in a post on the Azure Blog.

“Azure AI Studio also now integrates with Microsoft Purview (in preview), Microsoft’s service to prevent unauthorized access to data across apps and services, to discover potential ‘data risks’ in AI apps, impose encryption on sensitive data and govern AI app usage,” TechCrunch reports.

There are also new Studio tools to help prevent AI “jailbreaks,” the term for disabling model safeguards, as well as to detect hallucinations.

Microsoft has added the multimodal Phi-3-vision to the Phi-3 family of small language models announced last month. Microsoft developed the lightweight Phi family of SLMs, unveiled last year, for local use and things like smartphones.

Microsoft also announced the general availability of the multimodal GPT-4 Turbo with Vision on the Azure OpenAI Service.

SiliconANGLE draws attention to the new search functionality within Microsoft Azure AI Search, calling it “one of the most welcome updates for generative AI developers.” Microsoft’s retrieval-augmented generation (RAG) enterprise search solution comes a week after Google announced Gemini-powered search across all platforms.

OpenAI announced the availability of its new flagship model, GTP-4o, on May 13. TechCrunch says “the ‘o’ stands for ‘omni,’ referring to the model’s ability to handle text, speech, and video.” Mashable explains what GTP-4o can do on Azure.

In another splashy reveal, Microsoft said its pre-release Azure Virtual Machines featuring Cobalt 100 processors will be previewing to Azure subscribers who successfully apply for an invitation, according to The Register.

“Virtual machines are basically software-simulated versions of computers, so, instead of having to run 50 individual desktops, you can run 50 virtual machines on a high-powered server,” Yahoo explains.

Related:
Microsoft Azure AI Gains New LLMs, Governance Features, InfoWorld, 5/21/24
Microsoft Releases Upgrades to Azure AI Speech at Build 2024, ZDNet, 5/21/24
Microsoft Will Soon Start Enforcing Azure MFA Logins, TechRadar, 5/20/24