Microsoft Announces General Availability of 'Phi-Silica' Language Model for All Copilot+ PCs



At the keynote speech of Microsoft Build 2024, the company's annual developer conference held on May 21, 2024, Microsoft announced more than 40 APIs, including Phi-Silica, a small-scale language model that runs locally on the NPU installed in the AI-specialized PC

Copilot+ PC .

New models added to the Phi-3 family, available on Microsoft Azure | Microsoft Azure Blog
https://azure.microsoft.com/en-us/blog/new-models-added-to-the-phi-3-family-available-on-microsoft-azure/

NVIDIA Collaborates With Microsoft to Help Developers Build, Deploy AI Applications Faster | NVIDIA Blog
https://blogs.nvidia.com/blog/microsoft-build-optimized-ai-developers/

The keynote speech can be seen below.

Full Keynote: Satya Nadella at Microsoft Build 2024 - YouTube


In his keynote speech, Microsoft CEO Satya Nadella announced the Windows Copilot Library as a tool to make Windows the best platform for building AI applications. According to Nadella, the Windows Copilot Library is a library of local APIs that can be used immediately on Windows, and Phi-Silica is one of the AI models included in it.



Phi-Silica is one of the Microsoft language models in the 'Phi' family, and is the smallest model with 3.3 billion parameters. In addition, Phi-Silica is optimized for the NPU of the Copilot+ PC and can be run locally.

In addition, CEO Nadella announced three new models to the Phi family: Phi-3-vision, Phi-3-small, and Phi-3-midium.



Among these, Phi-3-vision is a multimodal AI model that can also recognize images, despite its compact size of only 4.2 billion parameters.



All three newly added Phi-3 models are available on Microsoft Azure.



Microsoft also announced that it has partnered with NVIDIA to provide three new models of Phi-3 for NVIDIA's inference microservice, NVIDIA NIM, all of which are optimized with NVIDIA TensorRT-LLM and are available at ai.nvidia.com or through NVIDIA AI Enterprise on Azure Marketplace.



Nadella also announced that support for OpenAI's

GPT-4o will be added to Azure AI Studio. During the keynote speech, he demonstrated how GPT-4o recognizes the screen of the game 'Minecraft' and plays the game while talking to it.



Additionally, Microsoft

announced an expanded partnership with online AI platform Hugging Face to bring text embedding reasoning directly to Azure AI Studio.

in Software,   Video, Posted by log1i_yk