Microsoft, this week at its Build 2023 conference, looked to press its generative artificial intelligence (AI) advantage further by previewing how these advances will be embedded both within the core Windows 11 operating system and its entire suite of applications.
A generative AI capability, dubbed Copilot for Windows, will, for example, make it possible to use natural language queries from the task bar to, for example, copy and paste data between applications, summarize documents and even plan a trip.
Similarly, Microsoft also showed how Microsoft 356 Copilot will enable sales and marketing professionals to automate a wide range of tedious tasks.
In addition. Microsoft showed how its Bing search engine will be integrated with ChatGPT to provide results to queries that are based on citations of more timely data. The general purpose Generative Pre-Trained Transformer (GPT) model that Microsoft licensed from Open AI was trained on data collected prior to January of 2022.
Microsoft CEO Satya Nadella told conference attendees these types of capabilities will dramatically increase productivity for all end users of its software by embedding intelligence reasoning with its software in a way that extends the capabilities of Microsoft Graph, an overlay that Microsoft employs to connect various application service via application programming interfaces (APIs). “This is going to make every user of Windows a power user,” said Nadella.
As part of that effort, Microsoft also showed how a plug-in framework will enable third-party application to take advantage of those same capabilities. There is already more than 50 plug-ins that take advantage of the generative AI models Microsoft makes available, with thousands more on the way, noted Nadella. That plugin framework is the same one that OpenAI previously made available.
Other AI capabilities that Microsoft previewed included Azure AI Studio, a development environment for training and managing AI models that enables organizations to “ground” their own data to further train generative AI models that Microsoft makes available. Those models consist of both the proprietary ones that Microsoft has created as well as open source models that Microsoft is making available via its Azure cloud service. Organizations can then create their own “Copilot Stack” to automate various tasks across a portfolio of custom applications.
Microsoft is also previewing a prompt flow tool that make it possible to more easily tune multiple models that might be employed across a workflow. Microsoft is also planning to make available a catalog for tracking models, modeling monitoring tools and dashboards through which data science teams can evaluate model errors and fairness issues.
In addition, Microsoft is now making generally available a language translation tool capable of converting documents created into more than 125 languages. Microsoft is also previewing an Azure Cognitive Search tool to make it easier to track models stored in Azure and Azure Cognitive Service for Language, a summarization capability that Microsoft is making available to application developers.
Microsoft is also moving to make it simpler to manage the volume of data required to train those AI models via Microsoft Fabric, a data lake that enables organizations to unify data and analytics using a cloud service based on Microsoft Power BI analytics tools and a data warehouse that is available in preview.
Finally, there is Azure AI Content Safety, a service that assigns a severity score to flagged content that is then shared with human moderators to further assess.
Nadella noted there has always been a symbiotic relationship between major technology advances and economic growth. However, generative AI is different than previous advances in the sense that these capabilities are being made available to everyone around the globe at the same time.
Of course, there’s still a lot more to come, all of which, for better or worse, is going to be highly disruptive. OpenAI is already previewing a next iteration of its platform that adds support for creating images. It’s still too early to determine just how profound an impact generative AI will have, but it’s already clear the world as it was once known is never going to be the same.