News

Microsoft Launches Azure OpenAI

• By Kurt Mackie
• January 17, 2023

Microsoft's Azure OpenAI service has reached the "general availability" commercial-release stage, the company announced this week. 

The Azure OpenAI service offers organizations access to artificial intelligence (AI) models. Those models include GPT 3.5 for natural language responses, Codex, used to suggest code for developers, and DALL-E 2, which composes images based on text descriptions.

Additionally, Azure OpenAI users will "soon" have access to ChatGPT, according to a Jan. 16 Twitter post by Satya Nadella, Microsoft's CEO. ChatGPT is a "fine-tuned version of GPT 3.5," that's been used for solutions like answering questions and summarizing text passages.

Testimonials
Customer testimonials for the Azure OpenAI service were listed in Microsoft's announcement.

CarMax has used the service to produce summaries of customer car reviews across 5,000 Web pages, which had an "80 percent editorial approval rate." Moveworks, a maker of conversational AI platforms, including an IT help desk solution, is using the Azure OpenAI service for its products.

KPMG used Microsoft's service to pull tax payment information together from various information siloes. Al Jazeera Digital used it for content "summarization and translation," plus content extraction and style observation.

Azure OpenAI Use in Microsoft Products
The Azure OpenAI service is currently being used in other Microsoft software products and services, including GitHub Copilot, Power BI and Microsoft Designer, per the announcement:

This [Azure OpenAI service] is the same production service that Microsoft uses to power its own products, including GitHub Copilot, an AI pair programmer that helps developers write better code, Power BI, which leverages GPT-3-powered natural language to automatically generate formulae and expressions, and the recently-announced Microsoft Designer, which helps creators build stunning content with natural language prompts.

Microsoft Designer, introduced in October and currently at the preview stage, uses DALL-E AI to generate images from text descriptions.

The ChatGPT natural language AI capability that will be coming soon to the Azure OpenAI service reportedly will be used to augment other Microsoft products as well.

ChatGPT was said to be coming to power Microsoft's Bing search service, plus Microsoft Office applications, such as Outlook, PowerPoint and Word, according to an article published by The Information (paywalled), citing unnamed sources. A summary of those claims can be found in this Thurrott.com article.

Microsoft's OpenAI Investments
Microsoft and OpenAI are partners. Microsoft invested $1 billion in OpenAI back in 2019 and OpenAI uses Microsoft Azure services to host its workloads.

Microsoft reportedly has been contemplating a further $10 billion investment in OpenAI, according to a Semafor.com article, which pointed to unnamed sources. The proposed deal was described as giving Microsoft access to 75 percent of OpenAI's profits.

OpenAI, which is an AI general intelligence research and development company, also offers its services directly. However, Microsoft contends that its Azure service offering adds the "security, reliability, compliance, data privacy and built-in Responsible AI capabilities" of its cloud platform to the mix. Additionally, organizations get "regional availability" with Microsoft's Azure service.

Developer Limits
While the Azure OpenAI service is deemed ready for production use by organizations, based on its general availability status, Microsoft's Azure OpenAI landing page still listed it as being at "preview" stage (at press time). Pricing for the Azure Open AI service is described as being "similar to Azure Cognitive Services pricing models."

Microsoft appears to be limiting who can sign up for the Azure OpenAI service, at least according to its preview description. Microsoft plans to vet the use of the service to ensure "responsible use," according to a FAQ on the service's landing page

The limited sign-ups are part of Microsoft's vetting process.

"As part of our Limited Access Framework, developers are required to apply for access, describing their intended use case or application before they are given access to the service," Microsoft's announcement explained.

Microsoft uses "content filters" to find hateful and abusive content. In such cases, developers may be asked to take immediate action.

Microsoft itself had early experience to that effect when an AI-based chatbot it created called "Tay" was abused by chatters to spout hateful rhetoric. Microsoft reacted by pulling Tay and apologizing back in 2016.