Advanced Formula Environment is becoming Excel Labs, a Microsoft Garage Project!

[Guest petraronald]

Experimentation is key to unlocking innovation in any field. It’s how we learn and grow, and it’s what helps us continue to create features that help you do more.

 

 

 

So today, we are thrilled to announce our continued investment in experimentation through the release of Excel Labs, a Microsoft Garage project. Excel Labs is an add-in that allows us to release experimental ideas for you to try, and to give us feedback that helps us evolve Excel to be most useful for you.

 

 

 

The Microsoft Garage is Microsoft’s official outlet for experimental projects across the company so that teams may receive early feedback from customers and better determine product market fit. With Excel Labs, in alignment with the Garage’s mission, expect to find very early-stage ideas that we are thinking about and wanting to evaluate with our customers. While some of these ideas may never make it to the Excel product, we believe that having this avenue to get your feedback will be crucial to creating more exciting features that transform what’s possible in Excel.

 

 

 

The first two experimental ideas to be released as part of Excel Labs are:

 

 

 

• Advanced formula environment
  

 

This advanced editor for complex formulas was

previously released as a separate Garage project, the

Advanced Formula Environment

add-in

. But with the desire to experiment across a wide variety of features, we’re upgrading the AFE and rolling all of its functionality into

Excel Labs,

so you won’t need multiple add-ins installed.

 

 

 

• LABS.GENERATIVEAI custom function
  

 

As you saw last month with the announcement of

Microsoft 365 Copilot

, we see so much potential in the power of generative AI. So, with this experimental add-in, you can test the benefits of generative AI directly from the Excel grid using the LABS.GENERATIVEAI custom function. This function allows you to send prompts from the Excel grid to a generative AI model and then return the results from the model back to your worksheet. Please note that this is not part of M365 Copilot but rather a place for you to play and experiment with generative AI today within the Garage framework.

 

 

How to get started?

 

 

 

 

It’s easy – just install the Excel Labs add-in through the Office Store. If you don’t see the add-in when you type Excel Labs into the Office Store search box, your version of Office may not meet the minimum system requirements.

 

 

 

After the add-in is installed or updated, select the Excel Labs button in the Home tab. This will open the Excel Lab’s feature gallery page where you can choose which feature you want to use. To get back to the feature gallery at any time, select the overflow menu in the navigation bar.

 

 

 

Excel Labs Feature Gallery

 

 

 

 

 

Advanced formula environment

 

 

 

 

The Advanced Formula Environment feature is a tool designed to help you more easily author, edit, and reuse complex formulas and LAMBDA functions. While the built-in Excel Name Manager lets you name and create complex formulas, the Advanced Formula Environment adds to that experience by providing capabilities typically found in modern code editors, such as IntelliSense, commenting, inline errors, auto formatting, and code collapse.

 

 

 

For more on this feature, visit the previous blog post announcing the release of the standalone add-in. Remember, you won’t need to download this add-in separately as the feature is now part of the Excel Labs add-in experience.

 

 

 

Advanced Formula Environment functionality

 

 

 

LABS.GENERATIVEAI custom function

 

 

 

 

Considering the recent excitement surrounding generative AI, we’re interested to see what types of Excel scenarios you have that would benefit from these AI models. So, to better help you experiment, we created the LABS.GENERATIVEAI custom function that enables you to prompt and display the response of OpenAI’s large language models directly into the grid.

 

 

 

LABS.GENERATIVEAI custom function

 

 

 

What is Generative AI?

 

 

 

 

Generative AI models are large-scale language models that use machine learning to generate digital content like human-like text, code, or even images. Given a prompt, the generative AI model returns a response that is generated by the model’s algorithms.

 

 

 

The opportunities for using generative AI are endless as it can generate responses in a variety of different formats. For the LABS.GENERATIVEAI custom function, we are specifically connecting to the OpenAI API for large language models. This includes models such as “gpt-3.5-turbo” (model that powers ChatGPT) and “text-davinci-003”. Some of the most common use cases for these models include text generation, text completion, summarization, classification, text transformation, and Q&A. For example, you can prompt it to parse out keywords in a survey response or you can ask it to analyze the sentiment of a table of tweets. In general, we encourage you to experiment with different prompts to see how being more general or more specific impacts the responses you get. Instead of “Write a poem about Excel”, try something like “Write a poem about Excel that mentions PivotTables”. If you’re interested, you can get started with the examples in the add-in or explore some common prompt examples on OpenAI’s website.

 

 

 

Given the nature of large language models, it’s important to understand that these models can produce nonsensical or even confident but false responses due to their predictive nature. We are dedicated to building responsible AI, and the LABS.GENERATIVEAI custom function effort was created with our AI principles and the Responsible AI Standard in mind. Additionally, the OpenAI Moderation endpoint was incorporated into the feature for automatic filtering of responses. However, it is essential to remember that there are still considerations to be considered when using the AI system, such as verifying results before making decisions.

 

 

 

Overall, the extent to what generative AI can do is still being explored, and this is your chance to experiment with what it can do for you in the context of Excel.

 

 

 

How to use the LABS.GENERATIVEAI function?

 

 

 

 

When you select the LABS.GENERATIVEAI feature in the feature gallery, the first thing you need to do is add your OpenAI API key. If you don’t already have an API key, you need to register for an OpenAI account and then request one.

 

 

 

Once you have your key, add it in the task pane. You only need to do this step once.

 

 

Configure your OpenAI API Key

 

 

 

Now, you're ready to try out the generative AI custom function!

 

Enter =LABS.GENERATIVEAI to access the function in the grid and enter your prompt as the input to the function.

 

 

 

 

 

 

Response Settings

 

 

 

 

Similar to OpenAI’s Playgound, in the Excel Labs task pane, you can also adjust the settings of the model’s responses by selecting the ‘Generative AI model settings’ button. This reveals different settings that affect the format of the responses. Two key settings you may want to experiment with are temperature and maximum output length. Temperature determines how consistent the returned responses will be. For instance, if you want to consistently get the same response from the model, you should set your temperature setting to 0. The other setting you may want to play around with is the Maximum Output Length setting which controls the maximum number of tokens returned in the response. Please be mindful that when you increase the number of tokens, you may hit Open AI’s rate limit more quickly.

 

 

 

LABS.GENERATIVEAI Response Settings

 

 

 

 

 

While you can adjust the settings in the task pane, you can also do so directly in the grid by using the function’s optional parameters. For instance, the following function produces a response with a temperature value of 1 and a maximum output length of 100 using the “text-curie-001” model.

 

 

 

=LABS.GENERATIVEAI("write me a poem about Excel", 1, 100, "text-curie-001")

 

 

 

 

 

To learn more about the settings, visit Models - OpenAI API.

 

 

 

We’re looking for feedback!

 

 

 

 

Excel Labs was born from a collaboration between Excel and Microsoft Research Cambridge teams. We fundamentally believe in empowering Excel users to do more and feel that it’s vital to continually expand a suite of tools that allow for experimentation and transformation of what’s possible in Excel under the umbrella of Excel Labs.

 

 

 

We’re actively seeking feedback on this Microsoft Garage project. We want to know what works, what doesn’t, and what you’d like to see next. Please share your feedback via the GitHub page or in the Microsoft Excel community hub.

 

Continue reading...