At its Build developers conference, Microsoft unveiled its first features in a customer product powered by GPT-3, the powerful natural language model developed by OpenAI, which will help users build apps without needing to know how to write computer code or formulas.
GPT-3 will be integrated in Microsoft Power Apps, the low code app development platform that helps everyone from people with little or no coding experience — so-called “citizen developers” — to professional developers with deep programming expertise build applications to improve business productivity or processes. This includes apps to review non-profit gift donations, manage travel during COVID-19 or reduce overtime required to maintain wind turbines.
For instance, the new AI-powered features will allow an employee building an e-commerce app to describe a programming goal using conversational language like “find products where the name starts with ‘kids.’” A fine-tuned GPT-3 model then offers choices for transforming the command into a Microsoft Power Fx formula, the open source programming language of the Power Platform, such as “Filter(‘BC Orders’ Left(‘Product Name’,4)=”Kids”).
It’s one of the first implementations showing how GPT-3, running on Microsoft Azure and powered by Azure Machine Learning and one of the first internal uses of its new managed endpoints capability, can solve real-world business needs on an enterprise scale, Microsoft said.
With new features powered by GPT-3, Microsoft Power Apps users can describe a programming goal in conversational language and have it automatically transformed into Power Fx code.
While Power Fx is built on Microsoft Excel, and therefore much easier to use than traditional coding languages, creating complex data queries can still be a steep learning curve, and these new features help lower that curve.
“Using an advanced AI model like this can help our low-code tools become even more widely available to an even bigger audience by truly becoming what we call no code,” said Charles Lamanna, corporate vice president for Microsoft’s low code application platform.
Built by OpenAI, an independent AI research and deployment company, GPT-3 is a massive natural language model that runs exclusively on Azure.
Through a partnership with OpenAI that aims to accelerate breakthroughs in AI — from jointly developing the first supercomputer on Azure that is powerful enough to meet the demands of very large AI models to testing and commercializing new AI technologies — Microsoft has a license to the code behind the GPT-3 model that allows it to integrate the technology directly into its products.
“This will allow people to query and explore data in ways they literally couldn’t do before, and that will be the magical moment,” Lamanna said.
Although these “citizen developers” didn’t need to know computer programming languages, they still previously had to understand the logic of writing formulas that might look something like this: FirstN(Sort(Search(‘BC Orders’, “stroller”, “aib_productname”), ‘Purchase Date’, Descending), 10).
With the new GPT-3-powered features, a person can get the same result by typing plainspoken language like: “Show 10 orders that have stroller in the product name and sort by purchase date with newest on the top.”
The features don’t replace the need for a person to understand the code they are implementing but are designed to assist people who are learning the Power Fx programming language and help them choose the right formulas to get the result they need. That can dramatically expand access to more advanced app building and more rapidly train people to use low code tools.
The new features announced at Microsoft Build will be available in preview in the English language throughout North America by the end of June.
Using an advanced AI model like this can help our low-code tools become even more widely available to an even bigger audience by truly becoming what we call no code.
“GPT-3 is the most powerful natural language processing model that we have in the market, so for us to be able to use it to help our customers is tremendous,” said Bryony Wolf, Power Apps product marketing manager. “This is really the first time you’re seeing in a mainstream consumer product the ability for customers to have their natural language transformed into code.”
GPT-3 is part of a new class of models, which Microsoft is broadly exploring through its AI at Scale initiative, that learn from examining billions of pages of publicly available text. They so deeply absorb nuances of language, grammar, knowledge concepts and context that the same model is able to perform a broad set of tasks that involve generating text.
OpenAI released an Azure-powered API last year that allows developers to explore GPT-3 capabilities. Since then, people have used it to do everything from writing poetry and tweets to generating articles, summarizing emails, answering trivia questions and generating computer code from plain language.
This discovery of GPT-3’s vast capabilities exploded the boundaries of what’s possible in natural language learning, said Eric Boyd, Microsoft corporate vice president for Azure AI. But there were still open questions about whether such a large and complex model could be deployed cost-effectively at scale to meet real-world business needs.
“We’re finding ways to bring it into Azure and our mainstream products,” Boyd said. “We think there are a whole bunch more things that GPT-3 is capable of doing. It’s a foundational new technology that lights up a ton of new possibilities, and this is sort of that first light coming into production,” he said.
The Power Platform team, which also works on low code tools to boost business productivity such as Power BI, Power Automate and Power Virtual Agents, quickly realized that GPT-3’s ability to translate conversational language into code could help advance the core mission of democratizing software development, or making it more straightforward for a wider variety of people.
We think there are a whole bunch more things that GPT-3 is capable of doing. It’s a foundational new technology that lights up a ton of new possibilities, and this is sort of that first light coming into production.
The goal is to have AI help with some of the more mundane elements of coding and formula expression, to both widen the pool of people who are able to use the tools and to free up experienced developers to focus on more interesting problems, like getting to the core of the business solution or building a beautiful interface.
Microsoft plans to infuse Power Fx into other tools within Power Platform, at which time the new natural language features powered by GPT-3 will expand into those products as well.
The Power Platform team worked closely with the Azure AI team to fine tune a GPT-3 model using Azure Machine Learning that could translate between natural language and Power Fx expressions.
The Power Platform team used Azure Machine Learning managed endpoints, a new capability announced in preview at Build that helps people deploy models of all sizes in Azure without needing to intricately manage underlying compute infrastructure. In one of the first internal use cases, the Microsoft product team is using it to deploy and manage the GPT-3 model that the team is using to offer new capabilities to Power Apps users.
The team also added filters to help detect sensitive or inappropriate content in any results that might get returned. The fact that the model in this circumstance is generating prescribed Power Fx formulas makes unintended outcomes less likely than, say, asking it to generate the answer to an open-ended question, Lamanna said.
And in the same way that you type a question into a search engine and then get to decide which result to click on, GPT-3 returns multiple suggestions for Power Fx formulas. The person building the app then chooses the most appropriate one to use.
“In all cases, there is a human in the loop,” Lamanna said. “This isn’t at all about replacing developers, it’s about finding the next 100 million developers in the world.”
Top image: Charles Lamanna, corporate vice president for Microsoft’s low code application platform. Photo by Dan DeLong for Microsoft.
Jennifer Langston writes about Microsoft research and innovation. Follow her on Twitter.
Leave a Reply