Microsoft researchers have unveiled a large language model with the ability to enhance the interaction with and understanding of spreadsheets has been introduced, marking a significant leap forward for AI’s ability to handle Excel files.
Outlined in a paper, entitled ‘SpeadsheetLLM: Encoding Spreadsheets for Large Language Models,’ the team looks to explore the complexities and challenges of applying AI to structured, logical and formulaic datasets.
SpreadsheetLLM promises the ability to combine these datasets with the already proven capabilities of large language models.
SpreadsheetLLM could make AI more efficient in your Excel sheet
According to the researchers, “SpreadsheetLLM is an approach for encoding spreadsheet contents into a format that can be used with large language models (LLMs) and allows these models to reason over spreadsheet contents.”
The research notes the widespread use of spreadsheets in various business operations, from simple data entry to complex financial modelling. However, despite the importance of these numerical files, existing LLMs have struggled to structure the data and formulas. By encoding the data in a package that LLMs can comprehend, Microsoft’s researchers hope to unlock the processing power of current generative AI tools that use existing LLMs within the Excel environment.
By unlocking the full potential of spreadsheet data, businesses using generative AI tools could improve their data-based decision-making and improve the efficiency of other administrative tasks.
Although SpreadsheetLLM is currently little more than a research project, the work to improve compatibility between Excel files and existing AI tools does at least signify the continually evolving power of AI.
Moreover, while the company has not yet committed to building such a tool on a wider scale, Microsoft’s multibillion-dollar investment in ChatGPT maker OpenAI and its subsequent rollout of generative AI functionalities across its office apps implies that the tech giant is committed to making AI more accessible.