X

Microsoft's Mini AI Is Built to Run on Your Phone, and That's a Key Difference

The tech giant is building lightweight AI that's more power-efficient and can run on PCs and phones.

Ian Sherr Contributor and Former Editor at Large / News
Ian Sherr (he/him/his) grew up in the San Francisco Bay Area, so he's always had a connection to the tech world. As an editor at large at CNET, he wrote about Apple, Microsoft, VR, video games and internet troubles. Aside from writing, he tinkers with tech at home, is a longtime fencer -- the kind with swords -- and began woodworking during the pandemic.
Ian Sherr
3 min read
Woman with headset and phone using AI

Microsoft announced a new AI designed for phones and PCs.

Qi Yang/Getty Images

Many artificial intelligence tools need an internet connection to a massive computing datacenter in order to work. But Microsoft on Tuesday announced a new AI designed for phones and PCs that doesn't require that connection.

The company says that its new AI, called Phi-3 Mini, can rival popular web-based AIs like OpenAI's ChatGPT-3.5. Microsoft told Reuters that its new AI can compete with other AIs that are 10 times more expensive to power and run. 

"Phi-3 is not slightly cheaper, it's dramatically cheaper," Sébastien Bubeck, Microsoft's vice president of GenAI research, told Reuters.

AI Atlas art badge tag

Microsoft's effort to build more efficient AIs is part of a trend in the tech world to squeeze the impressive capabilities of chatbots like ChatGPT, Meta's Llama, Google's Gemini and many more, into smaller apps that can run on our computers and phones. 

A smaller AI model running locally on a device likely won't have the same depth of knowledge as its large language model counterparts. But it could work more quickly, be more personalized and better protect your privacy.

Year of the AI PC

Most AIs until now have instead relied on sprawling power-hungry server farms, often powered by chips from AI industry darling Nvidia. While those AIs will likely continue to offer more features at ever-faster speeds, tech companies increasingly appear to be turning their attention to building capable, smaller AIs too.

Facebook, WhatsApp and Instagram parent company Meta last week announced a version of its Llama text chat AI that can run either over the internet or on mobile devices, with the help of specialized chips such as those made by Qualcomm. Google last year announced its Gemini Nano AI, which it added to its Pixel smartphone line. AMD, Intel and Nvidia have also announced efforts to build additional AI features into their chips.

Meanwhile, Apple is reportedly planning to announce AI capabilities designed to run on iPhones, iPads and Mac computers during its Worldwide Developers Conference in June.

(For hands-on CNET reviews of generative AI products including Gemini, ClaudeChatGPT and Microsoft Copilot, along with AI news, tips and explainers, see our new AI Atlas resource page.)

For its part, Microsoft and many of its partners have dubbed 2024 the year of the AI PC, in which they're rushing to add AI capabilities to their software and devices. Microsoft even made its first official change to the PC keyboard in 30 years, adding a dedicated key for its Copilot AI, opposite the Windows key it added in 1994.

It's unclear how Microsoft's Phi-3 Mini will show up in our daily lives, but the company speculated in a conversation with the Verge that it could help power custom apps for companies that may not have a lot of computing power to run more sophisticated AIs. 

Editors' note: CNET used an AI engine to help create several dozen stories, which are labeled accordingly. The note you're reading is attached to articles that deal substantively with the topic of AI but are created entirely by our expert editors and writers. For more, see our AI policy.