Microsoft integrates the GPT-3 language model in low-code platform power apps
At its Build developer conference, Microsoft demonstrated the first functions of a customer product that was developed using the natural language model GPT-3. In the future, GPT-3 should turn people without programming skills into developers.
GPT-3 integrates with Microsoft Power-Apps, the low-code app development platform that anyone can use to build applications. The manufacturer explicitly addresses people with little or no programming experience and calls them “Citizen Developers”. Likewise, professional developers with in-depth programming skills should benefit from the integration of the GPT-3 language model from OpenAI.
Contents
First examples prove suitability
There are already first examples of applications created with power apps, such as apps, at reviewing donations for non-profit organizations that Managing travel during the corona pandemic or the reduction of overtime, which are necessary for the maintenance of wind turbines, help.
The GPT-3 integration works in such a way that, for example, an employee who creates an e-commerce app describes a programming target in conversational language. It could be like this: “Find products whose name starts with ‘children’.” Now GPT-3 takes this command and converts it into a Power-Fx formula from the open source programming language of the Power platform.
GPT-3 from OpenAI, a company for AI research and use co-founded by Elon Musk at the time, runs in the Azure cloud and is powered by Azure machine learning. Power-Fx is based on Microsoft Excel and is therefore much easier to use than its conventional programming language. Nevertheless, complex data queries in particular can clearly benefit from the GPT-3 integration.
“Using an advanced AI model like this one can help make our low-code tools available to an even wider audience and really become what we do No code call ”, says Charles Lamanna, responsible manager for Microsoft’s low-code application platform.
In September 2020, Microsoft secured exclusive access to the OpenAI language model through a large-scale partnership. This mainly included access to the model’s source code. Other manufacturers are also allowed to use GPT-3, but only without access to its source code. In addition, there is a certain pre-determination for third-party providers because GPT-3 is only available in the Azure cloud.
Microsoft had announced that it would use its extended license to integrate GPT-3 into parts of its product portfolio. In autumn 2020, the manufacturer did not want to say which products this would affect. Now we know: Power-Apps gets the conversational approach first.
Easier, in fact, undeniably
Lamanna is certain, “It will allow people to query and explore data in ways that they literally didn’t have before. That will be the magical moment ”. Indeed, Lamanna has an apt example.
Natural language input of the formula: GPT-3 in power apps. (Animation: Microsoft)
With Power-Fx, users didn’t have to master a traditional programming language either. However, they had to adhere to a strict logic when formulating the formulas. It could look something like this:
FirstN (Sort (Search (‘BC Orders’, “stroller”, “aib_productname”), ‘Purchase Date’, Descending), 10).
With the new GPT-3-controlled functions, the same user would type instead of the formula above:
Show 10 orders that have strollers in the product name and sort by date of purchase, with the newest at the top.
It goes without saying that this is easier. Just like you can type a question into a search engine and then decide which result to click, GPT-3 provides several suggestions for Power-Fx formulas. The person who creates the app then chooses the most appropriate one.
Microsoft praises the democratization of software development
Power Apps does not hide the actual formula, but only offers it faster for selection. In this way, someone who has handled some projects with Power-Apps and GPT-3 will in the end be quite competent in dealing with Power-Fx. The learning curve drops drastically and the gain of knowledge accelerates. Microsoft calls it the “democratization of software development”.
Lamanna vehemently denies that technology could make developers superfluous: “It’s not about replacing developers at all, but about finding the next 100 million developers in the world.”
