metamorworks - stock.adobe.com
Microsoft incorporates AI foundation model in business apps
The tech giant made the Florence foundation model available to a wider audience through Azure Cognitive Services. The service also is now available in Teams and other applications.
Microsoft made its AI foundation model Florence available in preview for the public.
The tech giant revealed on March 7 that Florence, trained with text-to-image pairs, is now integrated as a production-ready computer vision service in Azure Cognitive Service for Vision, which provides computer vision capabilities such as analyzing images, reading text and detecting faces with prebuilt image tagging.
Microsoft introduced Florence in 2021 as a foundation model that expands representations across computer vision, changing scenes from coarse to fine and movement from static to dynamic.
With the availability of Florence in preview, Microsoft also revealed that Reddit will use its newly updated Vision Services to generate captions for images on the social media platform. In addition, the tech giant said it plans to apply Vision Services to Microsoft 365 apps suite of business productivity applications.
Microsoft strength
"It's really an example of how Microsoft because of their broad and successful reach into the enterprise for productivity applications can take something like an image processor and make it universally available across their entire suite," said Karl Freund, an analyst at Cambrian AI.
In the past several months, Microsoft has applied OpenAI's large language model across several applications, moving quickly in adopting foundation and large language models such as GPT compared to its chief rival in AI technology, Google.
"[Google] can't get out of their own way to get product to market and get out of research mode," Freund said.
Google has a more extensive search brand and more to lose if it were to take the same risks Microsoft takes in applying these experimental models to all its applications, Freund said.
Microsoft's integration of the foundation model into Azure Cognitive Services also benefits developers, said Arun Chandrasekaran, an analyst at Gartner. Since the services are available as an API, developers can easily consume them and embed them into whatever applications they are building.
Risks to mitigate
However, foundational models like these come with risks that enterprises must mitigate.
One of those hazards is presented by the training data, Chandrasekaran said.
"The training datasets for these AI services need to be free of copyrighted and other forms of proprietary content," he said.
Moreover, the output delivered by these services must be monitored through extensive pilots, Chandrasekaran added.
This will help users ensure more accuracy, be able to adapt how they use the foundation model for their own applications and preserve privacy, he said.
Nonetheless, AI foundation models such as Florence will appeal to enterprises in industries that use large amounts of images and videos, such as automotive, media, retail and manufacturing, Chandrasekaran said.
