I was on vacation in Europe during the Microsoft Build conference, so I did not watch it live. I returned to work last week and watched the two keynote addresses. Microsoft CEO Satya Nadella, CVP Scott Guthrie, and others provided information on the current state of Microsoft technologies, as well as a few significant announcements. As expected, Artificial Intelligence (AI) news dominated the keynotes.

Here are the announcements from the keynotes that caught my attention.

GPT-4o LLM

Microsoft introduced the GPT-4o Large Language Model and added this to their Azure OpenAI services and in Azure AI Studio.

This model is faster, less expensive, has higher rate limits, and supports more languages than GPT-4.

GitHub Copilot Extensions

I have been using GitHub Copilot (GHCP) extensively to help with coding documentation, writing tests, and other tasks. Microsoft has introduced some useful extensions to help in other areas. GitHub Copilot for Azure is an interesting one that allows you to query cloud resources from within Visual Studio Code. Other interesting extensions include ones for Docker, MongoDB, and Teams.

Soon, developers will be able to create their own extensions to GHCP.

More info

Small Language Models

The Phi-3 family of Small Language Models (SLMs) are faster, cheaper, and smaller (duh) than Large Language Models (LLMs) yet provide much of the same functionality. Phi-3-mini, Phi-3-small, and Phi-3-medium provide SLMs of increasing size depending on your needs. Because of their size, they are perfect for working locally using Edge computing scenarios. In addition, Phi-3-vision is a small model that includes image recognition and generation in addition to its language capabilities.

Azure OpenAI Studio is in General Availability

This is portal that allows you to build AI solutions, such as chatbots without quickly using a graphical interface. After months in preview, this service is now generally available.

CosmosDB Vector database capabilities

Azure Search uses the features of a vector database to perform semantic search queries. To access data in a CosmosDB database, we currently need to synchronize that data with a Vector database. Adding native Vector capabilities to CosmosDB for NoSQL eliminates the need for this synchronization, reducing latency and complexity. This feature is currently in preview.
More info

Improvements to Microsoft Dev Box

Microsoft Dev Box provides a cloud-based workstation to develop code.

Users and administrators will be able to customize DevBox for use by themselves and/or their team members, including tools and libraries required by the projects on which they are working. Admins will also be able to set up hibernation scale to save costs when a dev box is not in use.

More info

Partner Stories

The keynote included many stories and case studies from partners, but the most interesting one came from Khan Academy, which is using Azure AI tools to build Khanmigo, a service that provides AI services to teachers. They plan to provide Khanmigo free to all K-12 educators.

More

There were many more announcements - too many to list here. The Microsoft Build Book of News provides more announcements, as well as more details on the above announcements.