On the 12th, Microsoft released a new large language model, planning to develop a new AI language model - SpreadsheetLLM for spreadsheet applications such as Excel and Google Sheets.

Microsoft stated in a paper that SpreadsheetLLM, as a new AI model, will be widely used to understand and process complex spreadsheet data.

SpreadsheetLLM has the potential to change the way spreadsheet data is managed and analyzed, paving the way for more intelligent and efficient user interactions.

This may make accountants and data analysts worry about their future work prospects. Netizens on social platform X joked that "Karen's job will soon be replaced by artificial intelligence."

"Karen may soon be unemployed."

Researchers pointed out that current spreadsheet applications have rich functions and provide users with a variety of layout and formatting options, making it difficult for traditional AI language models to work in spreadsheet processing. SpreadsheetLLM is a AI model specially designed for spreadsheet applications.

Microsoft has also developed a SheetCompressor tool to help SpreadsheetLLM better understand and process spreadsheet data.

Researchers said that SpreadsheetLLM has a wide range of potential applications, from automatically performing daily data analysis tasks to providing intelligent insights and recommendations based on spreadsheet data. For example, SpreadsheetLLM can be used to automatically generate financial reports, identify anomalies or trends in data, and provide personalized product or service recommendations to customers.

Therefore, SpreadsheetLLM may completely change the way enterprises process data.

"As we know, LLMs that can write SQL will kill the entire data engineering industry."

"SaaS is in deep trouble."

"This will have a huge impact on the financial industry."

Ethan Mollick, associate professor at the Wharton School of Business, University of Pennsylvania, wrote on Twitter: "This once again shows that LLMs will soon be able to handle structured and unstructured spreadsheet data. This will unlock many use cases (predictions, finance, valuations, etc.), and having real sources of spreadsheet data often reduces illusions."

How does SpreadsheetLLM work?

By encoding spreadsheet data into a large language model (LLM) understandable, SpreadsheetLLM allows LLMs to reason about data, answer questions about it, and even generate new spreadsheets based on natural language prompts.

The core of SpreadsheetLLM is the "SheetCompressor" framework, which effectively compresses and encodes spreadsheet data to make it more easily processed by LLMs. SheetCompressor consists of three modules:

▲Structure Anchor-based Compression: Placing structure anchors throughout the entire spreadsheet to help LLMs understand the data structure.

▲Inverted Index Translation: Translating the spreadsheet into a more compact format and eliminating redundant data.

▲Data Format-Aware Aggregation: Grouping adjacent cells based on their numerical format and data type.

Illustration of the SHEETCOMPRESSOR framework (Image: Microsoft)

According to Microsoft, SpreadsheetLLM significantly improves performance in electronic spreadsheet detection tasks and outperforms regular methods by 25.6% in the context learning setting of GPT4. It reduces the cost of using tokens by 96% and provides better processing results.

Currently, Microsoft has not announced when it will release news about SpreadsheetLLM to the public. The paper pointed out that the model still has some limitations, such as limited understanding of complex or highly structured data, and SheetCompressor is currently unable to compress cells containing natural language.