The company appears to be pushing generative AI experiences from the cloud down into Windows devices; at the very least, that's what it's signalling it hopes to achieve with the roll-out of its new Windows AI Studio.

That is Windows AI Studio, unveiled today at the Microsoft Ignite 2023 conference and set to land in preview in the next few weeks. In effect, it's a kind of successor to the now-defunct AI Platform for Windows developers. Windows AI Studio amalgamates AI tools with a catalog of generative AI models that developers can fine-tune, customize, and deploy for local, offline use within their Windows apps.

But Windows AI Studio does not make developers deploy models offline — or in the cloud, for that matter. Instead, it gives them options to run models locally, or if that matters, in remote data centers, or in a hybrid local-cloud configuration.

"Given the pace of AI, we want to help developers quickly jumpstart AI development locally on Windows by giving them the tools and resources they need," says Logan Iyer, distinguished engineer in Microsoft's Windows systems and silicon integration group, via email interview with TechCrunch. "Our priority is to offer developers these tools as soon as they're available."

Windows AI Studio's model catalog accesses the models in Microsoft's and third-party repositories-on Hugging Face, for example. Developers can select, configure, fine-tune, and test models that are in an app or project locally, using their own data sets, before figuring out where they want to run those models with the new experience.

"We are … bringing together cutting-edge tools and a model catalog with models that we have tested, and in some cases, optimized for Windows so developers can jump-start AI development locally on Windows," Iyer said. "Today, developers lack a guided step-by-step interface to fine-tune their models — which is a blocker in deploying generative AI faster in their apps.". Windows AI Studio will provide this guided interface so developers can focus on writing code as we do the heavy lifting and help them bring gen AI features into their applications.
 
Later this year, we will roll out a feature called prompt flow, which will enable developers to have their apps switch automatically, as required, between smaller, faster models running on the local machine to larger, more capable cloud-hosted models. But on supported hardware-meaning hardware with either a dedicated AI accelerator chip or GPU-it will also pack more advanced models that run on the local device, such as Meta's Llama 2 text-generating model (thanks to a newly enlarged partnership with Meta), and the text-to-image model Stable Diffusion XL from Stability AI.
Also in the cards is Windows AI Studio as a VS Code extension-that's the VS Code open-source code editor.

Developers will have more choice of running their models either on the cloud … or natively on the edge, locally on Windows to meet their needs," said Iyer.

Given the challenges today around securing the necessary cloud resources to run AI models at scale, any tool that makes it easier to deploy models locally is likely to be well received by developers. How easy is Windows AI Studio with regard to running a performant model locally? Nobody knows until seeing a hands-on demo, and that will have to wait until the public preview to see the answer.