Microsoft launches lightweight AI model

Microsoft on Tuesday launched a lightweight artificial intelligence model, as it looks to attract a wider client base with cost-effective options.

The new version called Phi-3-mini is the first of the three small language models (SLM) to be released by the company, as it stakes its future on a technology that is expected to have a wide-ranging impact on the world and the way people work.

“Phi-3 is not slightly cheaper, it’s dramatically cheaper, we’re talking about a 10x cost difference compared to the other models out there with similar capabilities,” said Sébastien Bubeck, Microsoft’s vice president of GenAI research.

Adobe to bring full AI image generation to Photoshop this year. SLMs are designed to perform simpler tasks, making it easier for use by companies with limited resources, the company said.

Phi-3-mini will be available immediately on Microsoft cloud service platform Azure’s AI model catalog, machine learning model platform Hugging Face, and Ollama, a framework for running models on a local machine, the company said.

Last week, Microsoft invested $1.5 billion in UAEbased AI firm G42. It has also previously partnered with French startup Mistral AI to make their models available through its Azure cloud computing platform.

Meanwhile, Adobe said on Tuesday it plans to place a tool for full artificial intelligence image generation in its Photoshop software later this year.

Adobe’s image and video editing tools are widely used by creative professionals, but it faces rising competition from startups such as Microsoft-backed OpenAI, Midjourney and Stability AI, all of which offer services that can generate images from text prompts.

Adobe is developing its own image-generation AI system called Firefly, which is trained on data that Adobe has rights to, in order to avoid copyright infringement claims against users.

Adobe previously released image-generation tools in Photoshop that can fill in or expand parts an existing image. At a conference in London on Tuesday, the company said full image generation will come later this year, based on a new AI system called Firefly Image 3.

Much of Adobe’s focus has been on speeding up the work of professionals who use its software. The new image-generation tool will have the ability to tap a user’s uploaded image as a reference for the general composition of an image.

For example, a designer could make a quick sketch of a scene on a napkin, snap a photo of that napkin with a smartphone and then ask Photoshop to generate fully featured images in a variety of styles, said Ely Greenfield, chief technology officer for digital media at Adobe.

“Rather than having to very carefully describe exactly what goes where and try to make sure that I’m specifying the things I want things and that I don’t, it’s borrowing from the reference. So this is an amazingly powerful capability,” Greenfield said.

Adobe said a test “beta” version of the software is available to some users on Tuesday but did not give a date for general availability.

Earlier, Microsoft Corp will invest about $2.9 billion over the next two years to boost its artificial intelligence (AI) business in Japan, the Nikkei newspaper reported on Tuesday citing the company president.

The U.S. tech firm will announce the plans soon when Japanese Prime Minister Fumio Kishida visits the United States, the Nikkei said in its interview with Microsoft President Brad Smith.