Ever since Microsoft invested $1 billion in OpenAI in 2019, Microsoft officials have been working to find ways to integrate OpenAI's GPT-3 natural-language-model technology into various Microsoft products and services.
Microsoft secured a deal via which Azure became OpenAI's exclusive cloud provider. Microsoft created an Azure-based supercomputer to train OpenAI's massive AI models. In 2020, Microsoft obtained an exclusive license to GPT-3, though OpenAI still offers GPT-3 and other models via its own Azure-hosted programming interface. Microsoft also sells what it calls the Azure OpenAI service, (members) which adds Microsoft-provided enterprise capabilities to the core OpenAI models.
Microsoft already uses OpenAI features internally to power natural language processes in Power BI, Power Apps, and GitHub Copilot (members). Earlier this week, as originally reported by The Information and confirmed by Bloomberg, Microsoft also is working on a way to bring OpenAI's GPT technology to Bing via ChatGPT.
ChatGPT is a chat bot that is in public preview from OpenAI that isn't connected to web-search results. Instead, it is fine-tuned from the GPT-3.5 language model to product text, according to OpenAI. The ChatGPT models were trained on data from the Internet written by humans, including conversations, in order to make it sound more human-like.
Many are assuming that the big advantage Microsoft could gain from ChatGPT is in improving the way web-search results are rendered in Bing. But the real potential value of ChatGPT to Bing isn't mostly about natural-language answers to web-search queries, given that Bing, like Google, already has been providing for years more than a bunch of blue links in response to web-search queries.
Sure, ChatGPT could evolve Bing web search into something more conversational, rather than having users craft query terms to get the results they need. But do people really need that? It feels as if users have become accustomed to and proficient at crafting queries, and conversational search could be cumbersome. Knowledge workers and power users don't need the overhead of conversing with "intelligent" agents.
What if ChatGPT enables Microsoft to pull over the goal-line its long-rumored and occasionally seen Bing chatbot? ChatGPT also could play a role in providing more contextual answers not just to Bing web-search queries but also to unified Microsoft Search queries across more of Microsoft's enterprise products and services.
Microsoft has been working on a service called "Bing Concierge Bot" since at least 2016. For a while (during the rise and eventual fall of Microsoft's Cortana assistant), the Bing Concierge Bot project went quiet. But it reemerged in 2021 and has been tested publicly in various geographies since then in the form of a Bing chat bot. Some Bing users reported seeing the experimental Bing chat bot pop up when doing web searches at various points in 2021 and 2022.
Some may forget that Bing is more than just a web-search engine. Bing also is part of the Microsoft Search built-in search service for a number of Microsoft 365 enterprise products (members). If Microsoft ends up integrating ChatGPT into Microsoft Search, it potentially could bring deeper, more natural-language style responses to queries for business-specific queries within products like Teams, SharePoint and more, I'd think.
While it's easy to paint the OpenAI-Microsoft alliance as being a way for Microsoft to try to grow Bing's web-search share vs. Google, there are a number of other equally far-ranging implications for what ChatGPT and GPT in general could bring to Microsoft and customers.
Azure OpenAI service helps developers add advanced, pretrained language processing to applications with the usual Azure scaling and security capabilities.
Cognitive Services and AI Applied Services are collections of prebuilt AI components that help speed development, compared to building capabilities manually.
Azure Cognitive Services, a collection of machine learning cloud services, is removing some facial recognition features and restricting access to other features to support Microsoft’s responsible AI initiative.
Screenshots show how Microsoft’s AI Dashboard, in preview, can help identify potential errors and risk items in training data and models.
Sidebar explains how bias can creep into AI-enabled applications.
A new AI dashboard, in preview, can help organizations understand potential errors and bias in their models.
Azure Video Analyzer is being retired without leaving preview; possible migration destinations include Cognitive Services, Dynamics 365, or third-party solutions.
Several Microsoft offerings provide machine learning alternatives to Cognitive Services.
Azure Cognitive Services provides prebuilt machine learning configurations to help customers incorporate AI in their apps and processes.